You are on page 1of 11

A Modified Melting Model for Plastifying Extruders :'

D. R. HINRICHS and L. U. LILLELEHT,


Department of Chemical Engineering,
University of Virginia,
Charlottesville, Virginia

The extrusion melting model of Tadmor et al. (1967) has


been modified to include the effects of screw channel curva-
ture and melt leakage through the screw flight-barrel clearance.
An alternate coordinate system was used in modifying the
model to simplify the addition of curvature corrections. Also
proposed is an improved method of computing solid bed ve-
locity. A substantial number of typographical errors and trunca-
tions appeared in the original data published by Tadmor et al.,
arid verification of their computations with corrected data was
successful in all cases. Channel curvature corrections and the
improved method of computing solid bed velocity increased
the predicted rate of solid bed width decrease by approximate-
ly 14% during the initial stages of melting. Addition of leakage
flow to the model tended to offset this effect completely or in
part. Leakage flow effects always predominated during the
later stages of melting, however. These effects may interact
differently for other extruder sizes or screw designs; therefore,
the modified melting model should be a more accurate model
for use in the design of extruder screws and in studies of ex-
truder performance.

INTRODUCTION mercially available screws are equipped with special


he majority of thermoplastic conversion processes devices in the metering section to improve melt
T i n v o l v e the use of plastifying extruders in one uniformity, but the source of most erratic screw
form or another. The most common type of extruder behavior and wear seems to be the melting section.
utilizes a rotating screw having a helical channel of A more accurate understanding of the mechanisms
varying depth. A typical single-flighted screw, as involved in the melting process as well as operational
shown in Fig. 1, has a length 20 to 30 times its diam- and design factors affecting melting could ( a ) pro-
eter. The solid thermoplastic in granular, flake, or vide more efficient methods of operating existing
powder form is fed to the screw at one end; as it is screw designs, ( b ) facilitate better screw designs,
conveyed along the rotating screw, it is compacted, and ( c ) reduce the rate of screw wear.
melted, and mixed. Ideally, the plastic will reach Until several years ago, little was donc to de-
the discharge end of the screw as a homogeneous velop a usable model for the melting section of an
melt at a uniform temperature, pressure, and flow extruder. The practice has been to overdesign the
rate. feed and melting sections of the extruder screw and
The problem is that melt is not always homo- regulate output by a rigorously designed metering
geneous; neither are the temperature, pressure, and section (1). In a typical design method ( 2 ) it is
flow rate always uniform. The usual result is that simply assumed that the melting section can supply
process efficiency and product quality have to be molten plastic at a rate sufficient to keep the meter-
compromised. Output rates are usually reduced to ing section of the screw filled with melt.
improve product quality as determined by dimen- The earliest attempt to analyze the melting sec-
sional accuracy, surface uniformity, and physical tion of an extruder was by Maddock ( 3 ) . He postu-
properties. lated a melting mechanism that was confirmed by
Classically, the extruder screw is divided into Street (4)less than a year later. A second melting
three geometric sections as shown in Fig. 1: ( a ) the mechanism was proposed by McKelvey ( 5 ) based
feed section, ( b ) the transition or melting section, on superficial data obtained in experiments by Bern-
and ( c ) the metering section, although functionally, hart and himself ( 6 ) . In 1966, Tadmor ( 7 ) pub-
the sections are not exactly coincident. Many com- lished the first mathematical model based on Mad-
dock's proposed mechanism. A year later, Tadmor,
Paesented at the AIChE 67th National Meeting, held in Atlanta, Ga.,
Feb. 15-19, 1970. Duvdevani, and Klein (8) published a modified
268 POLYMER ENGINEERING AND SCIENCE, SEPTEMBER, 7970, Vol. 70, No. 5
A Modified Melting Model for Plastifying Extruders

melting model based on Tadmor’s original model. Solid polymer is continuously melted between the
Their predictions of the melting rate, as represented hot extruder barrel and the solid bed. Relative mo-
by the width of the unmelted solid bed along the tion of the extruder barrel drags the film of freshly
screw, showed reasonably good agreement with 19 melted polymer to the rear of the screw channel
sets of experimental data embracing five different where the trailing flight forces the freshly melted ma-
polymers. This model was also consistent with the terial to collect in a circulating melt pool. The solid
observations which led McKelvey to propose the bed deforms, as required, to fill the leading portion
second mechanism mentioned above. Another of the screw channel as the melting progresses. Heat
modification of the Tadmor Melting Model by is conducted from the hot barrel, through the melt
Chung ( 9 ) claims better agreement with experi- film, to the solid-melt interface where melting is ac-
ments; however, the technical validity of this deriva- tually taking place. Additional heat is supplied
tion is doubtful. through viscous dissipation in the melt film. Simul-
To date, the only unified approach to extruder taneously with the melting process, both the solid
performance based on mechanisms actually occurring bed and the molten polymer in the circulating melt
in all three sections of the extruder screw is pre- pool move toward the discharge end of the extruder
sented in a book edited by Klein and Marshall (10). due to the rotating, auger-like motion of the ex-
Using the Tadmor Melting Model, a theoretical in- truder screw. At some point along the screw, usually
vestigation of the effect of some operational and in the metering section, the solid bed disappears and
screw design parameters on melting in an extruder the melting process is complete.
screw has recently been published ( 11). One objective of this paper is to introduce correc-
The melting mechanism for both constant and tion factors into the Tadmor Melting Model to ac-
varying depth sections of the screw, as used by count for the true, helical nature of the screw chan-
Tadmor et al. ( 8 ) to develop their model for single- nel. The unwound-channel assumption used in all
flighted screws, is reviewed briefly below, previous melting models (7, 8, 9), although
The actual transition from solid polymer to melt is simplifying the mathematics, introduces several
seldom limited to the geometrically defined transi- errors into the model. The infinitesimal slice, dz, of
tion section in Fig. 1 and may begin in the feed sec- the channel in these models, shown in Fig. 3, cor-

FEED SECTION TRANSITION SECTION METERING SECTION vb


I “ I

\ /’”
PERPENDICULAR
TO F L I G H T
DEPTH DIAMETER

Fig. 1, Typical plastifying extruder screw. TRAILING


FLIGHT

TRAILIN
FLIGHT

ztdz
CIRCULATIN
MELT POOL

MELT

Fig. 2. Idealized channel cross section. POOL

tion and extend well into the metering section of the


screw which, geometrically defined, would contain - LEADING
only melt. Figure 2 shows a cross section of the ex- FLIGHT
truder screw channel at some intermediate stage of
the melting process where both solid and melted
polymer are present. The channel cross section \
LBOUN
B ED A
TWE E NR
S OY
LID BED
would actually be curved, but it is pictured here as AND C I R C U L A T I N G M E L T POOL
though the channel were unwound from the screw
and laid flat. F i g . 3. Mass balance diagram for channel axis.

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER, 1970, Vol. 70, No. 5 269
D . R. Hinrichs and L. U . Lilleleht

responds actually to the twisted, tapered slice of As shown in Fig. 3, the coordinate system of Tad-
Fig. 4. The actual solid bed velocity, width, and nior et al. was channel oriented. The unwound chan-
thickness of the infinitesimal slice depend on the nel was considered stationary while the barrel moved
distance from the screw's centerline. This shows, over it at the helix angle qb with velocity Vb. The
contrary to what has been assumed in previously x-axis was in the cross-channel direction, the y-axis
(not shown) in the channel-depth direction, and the
z-axis in the down-channel direction. Correction fac-
SCREW O R I E N T E D CHANNEL ORIENTED tors to relate the constant thickness, infinitesimal slice
COORDINATE COORDINATE
SYSTEM SYSTEM of Tadmor along the channel axis to the correspond-
ing twisted, tapered slice of the actual channel are
quite complicated to obtain. A cylindrical coordinate
system, instead of the Cartesian coordinates sug-
gested by the unwound channel, is used here to
eliminate most of these complications ( 12).
For this alternate approach, the 2'-direction is
along the screw axis, the r-direction is radial from
the screw centerline, and the &direction is circum-
ferentially normal to the radius. Planes defining an
. .
infinitesimal increment of the channel are normal to
I' z'tdz' the screw axis in this alternate approach as shown in
Fig. 4 . After the method of Tadmor et al., tempera-
F i g . 4. Coordinate system comparison for mass balance incre- ture and velocity profiles in the melt film are still
ment.
obtained by assuming the melt film is laid flat and is
of constant thickness at any chosen channel cross
published models, that the average solid bed width section.
and its width at the solid-melt interface are not the In Fig. 3, the j-direction is that of the velocity
same. vector in the melt film obtained by vectorial sub-
Determining the effect of leakage flow on the traction of the down-channel solid bed velocity,
melting rate will be the second objective of this V,,, from the barrel velocity, Vb.
paper. Typically, about ,005-in. clearance, increasing The polymer in the melt film is assumed to be a
with wear, exists between the extruder barrel and power law fluid described by:
the screw flight. As the screw rotates, some melt
could, therefore, leak from the circulating melt pool
at the rear of the channel through the screw flight-
barrel clearance and into the melt film at the front Assuming cross channel pressure effects to be
of the same helical channel one revolution upstream. negligible in the melt film, the equations of motion
In the case of multiple-flighted screws, melt would and energy reduce to:
actually be transferred from one channel to another.
This leakage flow tends to increase the thickness of = - K1 ( a constant)
rYj (2)
the melt film and reduce melting rate by ( a ) re- and
ducing the rate of viscous heat generation due to d2T d vj
km- - TYj -= 0 (3)
lower shear rates and ( b ) increasing resistance to
dY2 dY
heat transfer from the extruder barrel to the solid-
melt interface. These effects have not been con- These equations are coupled so, by analogy with
sidered in any of the previously published models. Newtonian flow, Tadmor et al. assumed a form of the
temperature profile:
THEORY
Development of the modified model below follows
T = A1t2 + (Tb - T m - A1 )t + Tm (4)
the basic method of Tadmor et al. (8). First, at an where A, is a constant yet to be evaluated by an
arbitrary cross section of the screw channel, melting iterative procedure.
is assumed to be a steady state process and an equa- The velocity profile in the melt film is obtained by
tion is developed which relates melting rate to the combining Eqs 1, 2, and 4, and integrating with the
width of the solid bed. Second, using the results of boundary conditions v j = 0 at t = 0 and v j = Viat
the first phase and assuming the velocity profile, [ = 1to give:
temperature profile, and polymer properties prevail-
ing at the particular channel cross section, an equa-
tion is derived to predict the dimensional change
of the solid bed along the screw axis. Because of this
assumption, the prediction of the complete melting
process must proceed incrementally along the length
of the screw using average properties and profiles
for each increment. where the symbol "erf" indicates the conventional
270 POLYMER ENGINEERING AND SCIENCE, SEPTEMBER, 1970, Vol. 10, No. 5
A Modified Melting Model for Plastifying Extruders

error function. The temperature profile in the melt Qe, per unit length along the screw axis is obtained
film is obtained by combining h q s 2, 3, and 5, and by integrating the circumferential velocity com-
integrating with the boundary conditions T = at ponent from E q 5 over the as-yet-unknown film
6 = U and 7' = T b at t = 1 as: thickness, and by applying the correction factor from
T - Tm K,v'ge i A a V A 9 E q 10 to give:
Tb - Tm = '+ 2( T b - Tm)A2km

The rate of heat transfer per unit area into the solid-
melt interface, -qyly=o, is obtained by differentiating
(6)
E q 6, evaluating it at the interface, and applying a
curvature correction factor, C,, to give:
where K , is a constant containing the melt film
thickness, 6. Using an iterative techiique, a value of
A, can be found such that Eqs 4 and 6 describe 6
nearly identical temperature profiles.
An estimate of the curvature correction, CQ,to the
velocity profile is obtained by assuming that the + 26A2 VA2
velocity profile in the melt film is distorted to the
same degree as it would be if the melt were New-
tonian and isothermal. The velocity profile for this
problem is given by Bird, Stewart, and Lightfoot
( 14). With appropriate change of notation, the total
tangential flow rate is obtained by integrating the
velocity profile between the solid-melt interface
diameter, Dm,and the barrel diameter, Db, as: Curvature of the channel increases the area of the
barrel with respect to the solid-melt interface. A
problem of this nature, involving pure conductive
heat transfer, has been solved by Bird, Stewart, and
Lightfoot (15) who show that the heat transfer area
Using the approximation ( 13) : should be based on the logarithmic mean diameter
of the two heat transfer surfaces. For the diameters
involved here, however, the logarithmic mean may
be very closely approximated by an arithmetic mean.
and since D , = D b -26, Eq. 7 may be written as: Assuming that the component of heat transfer due
to viscous heating in the melt film is affected by
vb8 ( 2 0 , - 36)Db channel curvature in the same way as the conduc-
Q=-. (9) tive component, the correction factor for heat trans-
2 2( Db - 6)'
fer, C,, may be expressed:
where VbS/2 is the Newtonian drag flow between a
moving and stationary flat plate. The curvature cor- (Mean diameter)
rection is then: of melt film 1 - D b - 8
c, = - (13)
Tangential flow rate (Solid-melt interface) Db - 26
( between cylinders ) - ( 2 D b - 3 6 ) D b diameter
cQ= (Flo;l:yeen
>-
2( Db - 6)*

(10)
Both E q 11 and E q 12 contain the unknown melt
film thickness, 6, which is obtained by relating Eqs 11
and 12 to each other through a mass balance on the
melt film and an energy balance on the solid-melt
The rate at which melt flows into the melt pool, interface.

POLYMER ENGINEERING AND SCMNCE, SEPTEMBER, 1970, Vol. 10, No. 5 27 1


D.R. Hinrichs and L. U . Lilleleht

The mass balance on the melt film may be written: sence of thermal effects induced by leakage flow.
Melting rate
[per unit area] =
of interface
+[ Melt flow
rate into
melt pool
1 Assuming an infinitely thick solid bed (justified by
the low thermal conductivity of polymers) with
heat conduction only in the y-direction, Tadmor
( 7 ) showed that a steady, exponential temperature
profile exists in the solid bed with respect to the
[Melt
rate through
leakage] solid-melt interface. Thus, he showed that heat re-
quired to melt a portion of the solid bed is the sen-
Xe clearance sible heat required to raise it from the initial solid
(14) bed temperature, T,, to the melting temperature,
T,, plus its latent heat of fusion, A. (Tadmor
where X B is the circumferential width of the solid-
idealized the actual heat capacity versus tempera-
melt interface. The velocity profile of the melt in the
ture curves to yield constant specific heats for the
clearance between the barrel and screw flight must solid and melt and to obtain discrete values for melt-
be known before the leakage rate can be determined. ing temperature and latent heat of fusion.) TO
Analytically, this profile would be practically im-
account for heat necessary to raise freshly melted
possible to obtain due to transient temperature ef- material to the average temperature of the melt en-
fects and unavailable boundary conditions. To cir- tering the circulating melt pool, Tadmor et al. (8)
cumvent this complication, it is assumed that the reduced the heat available for melting at the solid-
deviation of the velocity profile in the clearance from melt interface by the sensible heat necessary to
linearity has the same effect on leakage flow rate as raise the melt from the melting temperature to the
nonlinearity of the velocity profile in the melt film
has on flow rate in the melt film. With this assump- average melt temperature,T This is expressed:
tion, the volumetric flow rate of melt through the Heat per unit mass
clearance, QEO, is obtained by analogy with E q 11as: ( required for melting )
= c,, ( T , - T,) + A + c,, (T- T m ) (19)
To evaluate Cr, the temperature profile in the
where melt leaving the clearance and entering the melt
film must be known. As in the case of the velocity
profile in the clearance discussed previously, the
temperature profile would be nearly impossible to
determine for the same reasons. A reasonable esti-
mate is used instead.
The melt entering the clearance has just passed
over the circulating melt pool from the melt film of
the same channel (or from the identical, adjacent
channel for multiple-flighted screws). In so doing,
the melt is in contact with the barrel and is strongly
affected by the barrel temperature. The temperature
From Eqs 11, 15, and 16, the mass balance of E q of the melt is also influenced by adjacent fresh melt
14 becomes : which will be forced into the circulating melt pool
by the advancing flight. A reasonable assumption is
that melt entering the clearance will have a mean
(6-c) temperature corresponding to the average of the
barrel temperature and the mean temperature of the
(17) melt flowing into the circulating melt pool from the
melt film. Also, contrary to the original simplifying
Thermal effects will be associated with the leakage assumption of constant melt film thickness in a chan-
flow which will tend to either increase or decrease nel increment, the melt film is actually thinnest at
the mass rate of melting by an amount C,. The en- the point where leakage flow enters the melt film
ergy balance at the solid-melt interface per unit and is strongly affected by the leakage flow tem-
Y-axis length may be written: perature. A second reasonable assumption, there-
fore, is that excess heat carried by the leakage flow
Normal rate of Excess heat transfer rate will melt and increase the temperature of an equiva-
heat transfer ) (
-k due to leakage flow lent amount of the solid bed. With these assumptions,
(18) an estimate of the correction factor, C,, is expressed:
Heat required per
unit mass of melting (ofmelting
where the normal rate of heat transfer is considered Using Eqs 12, 17, and 19, the energy balance
as that through the solid-melt interface in the ab- given by E q 18 becomes:
272 POLYMER ENGlNEERING AND SCIENCE, SEPTEMBER, 1970, Vol. 10, No. 5
A Modified Melting Model for Plastifying Extruders

x’-axis length obtained by multiplying E q 15 by the


melt density.
The average temperature of the melt entering the
melt pool is obtained by integrating over the melt
film thickness:

The velocity, uj, is substituted from E q 5 and the


temperature from E q 4. The resulting expression is
lengthy and is given in Refs 8 and 12.
This completes the first phase of the development
which characterizes the melt film at any particular
+-)A3
d A2
[ e r f ( +z+-
d A2
) A3
channel cross section. The second phase of the de-
velopment shows how the solid bed varies along the
2‘-axis.
A mass balance on an infinitesimal increment of

+_-> A3

VA2
-.-I} A32
A2
(22)
the solid bed normal to the 2’-axis as shown in Fig. 4
is used to determine how the solid bed width varies
as a function of local melting rate. Assumptions in-
volved in this mass balance are that the melt film is
fully established and that the solid bed moves toward
Solving E q 21 for the melt film thickness, one ob- the discharge end of the extruder at a constant
tains : velocity, V,,,. The second assumption was verified
experimentally by Tadmor (7). This mass balance
CC is expressed as:

c, The solid bed velocity is established by dividing


+ pmVbCQU2
the volumetric flow rate of solid polymer at the start
[2kn ( T b - T m ) full xO of melting by the cross sectional area of the channel.
Cps (Tm - Tf) ++ Cpm (T - Tm)
(23) In the manuscript (16) of Ref 8, Tadmor et al.
proposed finding the cross sectional area normal to
the down-channel direction, A,,, according to the
Comparison of E q 23 with the corresponding, un- relationship:
corrected expression for the melt film thickness in
Ref 8 shows another implicit correction is contained A,, = H,( P COSY - W,) (28)
in E q 23. In the reference, the solid bed width is not where (pis the average of helix angles at the barrel
taken at the same helix angle as the barrel velocity and bottom of the channel. Use of an average helix
component. This error is attributed to the twisted angle introduces an error since the helix angle, q,
channel cross section when the channel oriented co- varies nonlinearly with diameter, D, in the channel
ordinate system is used. as :
D
The mass rate of melting, 00, per unit length along
the z’-axis is obtained by combining Eqs 17 and 23
to give:
An additional error may be introduced in determin-
ing at which level in the channel the average down-
channel velocity occurs since this point must be
used to determine the solid bed velocity at the
solid-melt interface which, in turn, is used to de-
termine V j . These errors seem to be a consequence
of using the channel oriented coordinate system
where both magnitude and direction of solid bed
velocity in the down-channel direction vary with
level in the channel as shown in Fig. 4.
With the screw axis oriented coordinate system,
the solid bed velocity along the screw axis is a true
constant and independent of diameter as shown in
Fig. 4. The circumferential width of the channel,

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER, 1970, Vol. 10, No. 5 273
D.R. Hinrichs and L. U.Lilleleht

WB1,at an arbitrary diameter, D1, assuming the flight For a tapered channel, the solid bed thickness will
thickness is constant, is: vary over the interval of integration of E q 27 as:

We1 = P D -
~ -w, (30)
H=H0-h‘ (35)
sin ‘pl Substituting Eqs 24, 33, and 35 into E q 27 and in-
tegrating by separation of variables using the bound-
where q1is the helix angle at D,. The cross sectional ary condition X B= XBl at z‘ = 0, the following ex-
area of the channel for the infinitesimal, circumfer- pression is obtained:
ential increment is obtained by integrating E q 30
between the diameter at the bottom of the channel,
Dh, and the barrel diameter, Db. In so doing, the solid
bed velocity is determined as:
XI3 f K5 - (K6
- (K6
K4xB)”

KJBI)’” 3
2(KO + K4Xo)” - I(4 - K7

* [ 2(K6 + K4X,)” - K4 + K7
(36)
- K4

2(K6 + K ~ X O ~ )-’ ’ ~ + - HO
1
K4 K7 KT

P
2(K6 + K4XB1)”’ - K4 - K7 -H,-Az‘
PDb --
sin ‘pb where

I 7rDh --
P
sin $oh.

The solid bed velocity to be used in evaluating Vj is


determined from:
VW
V,,linterface = - (32)
sin ‘pm
and
A correction factor, C,, assumed constant over the
range of integration of E q 27, is introduced to relate
K , = (K4 - 4K4K5 + 4K6)” ( 37d)
the average solid bed width to its width at the solid- For a constant depth channel, the solid bed is as-
melt interface : sumed to be of thickness HI over the range of inte-
gration of E q 27. Substituting E q s 24 and 33 into
XB E q 27 and integrating using the same boundary con-
C,=- (33)
x, dition as for E q 36, the following expression is ob-
tained:
To evaluate C,, it is assumed that it corresponds ex-
actly with the ratio of circumferential channel width ( KIII + KJ,, ) ”’- ( KIOf K8XB) ” Kg
at the solid-melt interface to the average circum-
ferential channel width below this radius. C, is ob-
( + ”’ -
KIO K8XBl) Kg
(38)
tained by evaluating E q 30 at the solid-melt inter-
face diameter, D,, and dividing by the average where
channel width. The average channel width is ob-
tained by integrating E q 30 between the diameter (39a)
at the bottom of the channel, Dh, and the solid-melt
interface diameter, and dividing by the thickness of
the solid bed. The correction factor then becomes:
and

Neither E q 36 nor 38 can be solved explicitly for


X,, thus necessitating the use of a numerical iteration
technique such as the Newton-Raphson scheme (17)
with objective functions defined by subtracting the
left sides of the respective equations from their right

274
! a +- P
sin q h
sides. Convergence is rapid: only three or four
iterations are usually required for accuracy to six
significant figures.

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER, 1970, Vol. 10, No. 5


A Modified Melting Model for Plastifying Extruders

Equation 27 was integrated assuming a, C , CE, of the melting model, the iterative solution was
and wee were constant, thus necessitating successive terminated in each increment along the screw when
application of the appropriate equation to adjacent, three conditions were met by successive iterations.
small increments along the screw using average These were: ( a ) the temperature change at the cen-
values of the constants of each increment. The com-. ter of the melt film was less than O.l"F, ( b ) the
putational scheme suggested by Tadmor et al. (8) melt film thickness changed less than 0.0001%, and
may still be used with the modifications noted ( c ) the solid bed width change was less than 0.0001
above. in. at the end of the increment. Equations resulting
Due to precision errors in evaluating the dif- from the assumption of a linear temperature profile
ference between error functions when the tempera- in the melt film were used whenever the other set of
ture profile in the melt film is nearly linear, Tadmor equations predicted a deviation from linearity at the
cat al. had developed a second set of equations to center of the melt film of less than 1.75"F. An ad-
characterize the melt film by assuming a linear tem- ditional switching criterion, not suggested in Ref 8,
perature profile corresponding to E q 4. The form was also used. Whenever any of the error function
of the final equations presented in this paper is not arguments involved in the computations exceeded
affected if the expressions for U1, and U2 given an absolute value of three, the equations for the
in Ref 8 for a linear ternperature profile are used in linear temperature profile assumption were used.
place of those given above, Prior to using any of the modifications presented
above, verification of the published computations
was attempted. After data given in Ref 8 was cor-
DISCUSSION rected for typographical errors and truncations (18),
In all cases discussed, the data of Tadmor et al. verification was successful in all cases to within the
(8, 18) were used. No other literature source of thickness of a line on a plot of solid bed width ver-
data of sufficient scope for the computations could sus position along the screw. The dotted lines in Figs.
be found. Published results and data points shown in .5 and 6 show typical verification results for low-den-
the figures to be discussed below were taken from sity polyethylene which coincide exactly with results
Ref 8 and its manuscript (16). For the computation published in Ref 8. The dashed lines in Figs. 7

Fig. 5. Leakage flow and curva-


ture effects for test N o . 1 . Poly-
mer type: low-density polyethy-
lene. Operating conditions: T b
3 5 7 9 II 13 IS 11 19 11 23 25 17 = 450°F; N = 40 rpm; G =
D I S T A N C E FROM S T A R T OF S C R E W , DIAMETERS 87.7 lb/hr.

Fig. 6. Leakage flow and curva- :


ture effects for test N o . 2. Poly- 0.9

mer type: low-density polyethy- f 0.8

h e . Operating conditions: T b 0.1

= 300°F; N = 60 T F ; G = 0.6

136.1 lb/hr.

3 5 7 9 11 I3 IS I7 I9 11 13 15 21

DISTANCE FROM START OF SCREW, DIAMETERS

I- 0.8 -
r 0.1 .
0.6 -
0.5 -
D
-
-
1
0
0.4

0.3 -- MODIFIED MODEL R E S U L T S


', -
Fig. 7. Leakage flow and curua-
ture efects for test No. 3 . P d y -
0.1
0.1. *>O
UNMODIFIED MODEL

DATA FROM DUPLICATE


RESULTS

RUNS
* * o . ~
'.. mer type: polypropylene. Oper-
: 0.0
0
' ' ' * I'I * ' ;I ' * ' 2'1 ' 2mating7
conditions: T b = 445°F;

D I S T A N C E FROM S T A R T OF S C R E W , DIAMETERS
N = 60 ~pm;G = 96.8 lb/hr.
275
POLYMER ENGINEERING AND SCIENCF, SEPTEMBER, 1970, Vol. 10, No. 5
D.R . Hinrichs and L. U . Lilleleht

thmugh 10 are also verification results and coincide


with results in the reference. The same computational
scheme was used to produce the dashed lines in
ke, *a.
1,-

Figs. 5 and 6 except more accurate viscosity data 0.6 --?l


(IS) were used.
The solid lines in Figs. 5-10 show typical results -MODIFIED MODEL RESULTS
UNMODIFIED MODEL RESULTS
of the modified melting model using the same data 0.2
- 0.Q DATA FROM DUPLICATE RUNS
o,l
used for the dashed lines. In all cases, the net effect 0
z o . 0 ' a a 1 ' " a " "

of the modifications was to predict a nearly identical 3 5 7 P II 13 IS I7 IP 21 13 25 27

DISTANCE FROM START OF SCREW. DIAMETERS


to a substantially faster melting rate during the
initial stages of melting as indicated by a more rapid Fig. 8. Leakage flow and curvature efects for test NO. 4 .
reduction of the solid bed width. In most cases, Polymer type: high-density polyethylene. Operating condi-
tiom: T b = 425°F; N z 60 'rpm; G = 122.1 lb/hr.
agreement with the data is approximately the same
as before the modifications were included in the
model. Data points on the horizontal axis of Figs.
5-10 indicate a broken or otherwise nonexistent I
+ 0.8
solid bed. Discounting these points, some portions
t
0
S 0.7
of the modified model curves are closer to the re-
maining data points than the corresponding portions
0 0.6

0.5
t 1
0

of the results of Tadmor et al. For other portions :0 . 4


0
0.3
of the curves, the opposite is true. This is undoubt- v)

: 0.2
edly a consequence of the particular combination u
= 0.1

of extrusion variables used for the tests such as :0.0 3 S 1 P I1 I3 IS 17 I9 21 23 25


screw design parameters and clearance with the DISTANCE FROM START O F SCREW, DIAMETERS

barrel (from Ref 18, the clearance was 0.0049 in.). Fig. 9. Leakage flow atid curuature efects for test- No: 5.
For the particular combination of parameters in- Polymer type: rigid polyoinyl chloride. Operating conditions:
volved, the effects of the essentially independent T I , = 375OF; N = 30 rpm; G = 107.2 lb/hr.
modifications tended to cancel. These effects are
discussed qualitatively below.
The use of the screw axis oriented coordinate
system introduced a curvature correction as dis-
cussed below E q 23. This correction, in effect, ac-

-
counts for the slightly different helix angles at the 0.5
barrel and at the solid-melt interface. The decreas- :0 4 -
ing helix angle as diameter increases will result in 0
0 3 -- UNMOOlFlED MODEL RESULTS

the prediction of a thicker melt film and, therefore, "z o 2 MODIFIED MODEL RESULTS

a faster melting rate. From the computations, melt :O I - .SO DATA FROM DUPLICATE RUNS
YI
= D o "
" " ' a ' " ' 8 ' a

film thickness is usually less than 0.020 in., and


E q 25 will be less than 1% larger than the corre-
sponding equation in Ref 8. This amounts to less
than 0.5% increase in predicted melting rate. This
is the only direct effect of using the alternate, screw
axis oriented coordinate system.
The factor, C,, relating average solid bed width factors have increased the predicted melting rate
to its width at the solid-melt interface will result in by less than 0.5% and 0.2576, respectively. Their
a faster rate of solid bed width decrease as can be effect would be greater for a smaller diameter screw.
seen by combining Eqs 24, 27, and 33. An upper Comparing the method of computing solid bed
limit on the value of C, may be estimated by the velocity by Tadmor et al. with the method using
ratio of barrel diameter to the average solid bed E q 31 shows that the latter will predict a 3% lower
diameter at a point in the screw where the channel velocity for the screw described in Ref 8. This, in
is deepest. For the screw described in Ref 8 and itself, will have no direct effect on melting rate pre-
used for these tests, this correction may increase the dictions; however, it will cause a 3% faster predic-
predicted rate of solid bed width decrease by 10% tion of the rate of solid bed width decrease,
during the initial stages of melting. The percentage The difference in solid bed velocity from using
would decrease as the channel became shallower the two methods will also produce a small difference
during the later stages of melting. in the shear rates occurring in the melt film. De-
From Eqs 10 and 13, it can be seen that the COT- pending on the relative screw speed and solid bed
rection factors to the velocity and temperature velocity, Vj will be, typically, less than 1% higher
profiles, CQ and C,, will both be greater than unity. which, in turn, would result in the prediction of a
For a melt film thickness of 0.020 in., CQand C, will correspondingly thinner melt film to sustain the
be greater than unity by 1%and 0.5%, respectively. prevailing melting rate. Ignoring other effects, the
Inspection of Eqs 24 and 25 shows these correction thinner melt film would increase heat transfer rate
276 POLYMER ENGINEERING AND SCIENCE, SEPTEMBER, 1970, Vol. 10, No. 5
A Modified Melting Model for Plastifying Extruders

from the barrel to the solid-melt interface. The over- Nylon 66 (Fig. l o ) , the photographs show thin wisps
all effect would be less than 0.5% faster melting of solid bed appearing periodically over the melt
rate prediction. pool. This is indicated by the cyclic nature of the
To this point, all of the modifications that have data points in the figure since the thin wisps were
been discussed have resulted in prediction of a faster included in the solid bed width measurements. En-
rate of solid bed width decrease. The total effect trainment of these thin wisps of solid bed by the
is estimated to be approximately 14% during the circulating melt pool would increase the apparent
initial stages of melting, decreasing as melting melting rate. Significant fracturing of the solid bed
progresses. as indicated by data points on the lower horizontal
The leakage flow modification acts in the opposite axis of Figs. 5-10 would also tend to invalidate the
direction: that is, melting rate prediction is signifi- melting models beyond the points of fracture since
cantly decreased by leakage flow. Equation 23 fracture indicates an abrupt change in solid bed
shows that the melt film thickness is increased by velocity.
at least one-half the clearance. Both heat flux from A more complete and detailed discussion of these
the barrel to the solid-melt interface due to pure and other anomalies may be found in Ref 12.
conduction and the viscous dissipation in the melt CONCLUSIONS
film would be reduced. Equation 24 shows that
melting rate is reduced by up to one-half the leak- The incorporation of curvature corrections into
age flow rate, relative to the melting rate without the melting model of Tadmor et al. was considerably
leakage flow. This effect would become more pro- simplified by using a screw oriented coordinate sys-
nounced as the solid bed width decreases. In Figs. tem instead of a channel oriented system. The two
5-10, relative to the unmodified model, the predicted most significant modifications were for leakage flow
melting rate becomes increasingly slower during and the correction factor relating average solid bed
the later stages of melting. This is most clearly seen width to its width at the solid-melt interface. All
in Fig. 6 as the upswept nature of the solid curve, modifications with the exception of leakage flow
and is due to the increasing influence of leakage flow tended to predict a faster rate of solid bed width
combined with the decreasing influence of the decrease. For the experimental data used, leakage
curvature corrections. Based on the computations flow effects ranged from greater than to less than the
for Figs. 5-10, leakage flow thermal effects given by effect of the curvature corrections during the initial
E q 20 are estimated to offset less than 10% of the stages of melting. During the later stages of melting,
other leakage flow effects discussed above. These leakage flow effectspredominated.
figures show that the net leakage flow effect may be The modifications discussed above should make
greater than (Fig. l o ) , equal to (Figs. 6 and 7 ) , or the melting model more relevent to the actual ex-
less than other effects for these tests during the ini- trusion process. Unfortunately, for the experimental
tial stages of melting. During the later stages of data available to test these modifications, the various
melting, leakage flow effects always predominated effects tended to cancel. This would probably not
and the dashed curves, representing melting rate be the case with different size extruders or screw
with no modifications, are steeper (in the negative designs. Unlike previously published models, the
sense). In all cases, as the effects of reduced melting modified melting model presented herein is capable
rate accumulate, the solid curves eventually rise of accounting for different screw flight-barrel clear-
above the dashed curves. ances and different curvatures of the screw channel.
Therefore, it should be a more accurate model for
Photographic data given in Ref 8 indicate that use in the design of extruder screws and in studies
the melting models developed to date may not be of extruder performance.
completely adequate for the experiments represented
by Figs. 7-10 since additional mechanisms may be ACKNOWLEDGMENT
occurring in the extruder that are not included in One of the authors (DRH) gratefully acknowl-
the models. For Figs. 7 and 8, the photographs in- edges a National Science Foundation Traineeship
dicate air or gas pockets between the solid bed and which made his part in this work possible. Also, the
circulating melt pool over most of the screw's transi- authors thank Dr. Z. Tadmor for providing more
tion section. Although these may have formed due accurate data and other helpful comments concern-
to shrinkage during cooling prior to extracting the ing this paper.
screw from the extruder, their size, shape, and loca-
tion also suggest a starvation condition in the ex- NOMENCLATURE
truder. Actual melting rates may be increased due
to the gas pockets permitting pieces to break away a = constant defined in E q I , OF-'.

A solid bed thickness taper per unit length


from the solid bed and become entrained by the
circulating melt pool. The experiment using poly- of screw.
vinyl chloride (Fig. 9 ) was the only one to use dry- A1 = constant defined in E q 3, O F - ' .
blended resin (the others used pellets). Entrainment A2 = aA,/n, dimensionless.
of resin particles in the melt film would produce a A3 = a( T , - T , - A,)/2n,dimensionless.
more rapid solid bed width decrease than would A,, = cross sectional area of solid bed normal
occur by melting alone. For the experiment using to x-axis, in.2

POLYMER ENGiNEERlNG A N D SCIENCE, SEPTEMBER, 1970, Yo!. 10, No. 5 277


D.R. Hinrich and L. U.Lilleleht

specific heat of melt and solid, BTU/lb- face at beginning of increment circum-
O F . ferentially, in.
curvature correction factors defined by Y = rectangular coordinate, channel depth
E q s 13,10, and 33. direction, in.
melting rate due to leakage flow thermal z = rectangular coordinate, down-channel
effects, lb/sec-in. direction, in.
diameter: D1,at arbitrary point; Db, at z’ = cylindrical coordinate. screw axis direc-
barrel; Dh, at bottom of channel; Dm, at tion, in.
solid-melt interface, in. 6 = melt film thickness, in.
mass flow rate through extruder, lb/sec. E = screw flight-barrel clearance, in.
solid bed thickness: H,, at beginning of 0 = cylindrical coordinate, circumferential
varying-thickness increment; Hi, in con- direction, degrees.
stant thickness increment, in. x = heat of fusion, BTU/lb.
thermal conductivity of melt, BTU/sec- t = y/s, reduced position variable, dimen-
in.-’ F. sionless.
constant defined by E q 2, 1b,/h2 pm, ps = density of melt and solid, l b / i r ~ . ~
( Kl/m,)l/n, sec-l. r?4j
= shear stress in j-direction, 1bf/in.’
moS2K2n+1, lbf/sec. Qr = expression defined by E q 25, lb/sec-
integration constants defined by Eqs
37a, b, c, and d; K4, K,, K,, in.; K6, in.’ cp = helix angle in channel: (p, average of
integration constants defined by Eqs those at barrel and bottom; ‘pl, at arbi-
trary point; ‘pb, at barrel; ‘ph, at bottom;
39a, b, and c; K,, K9, in.; Klo, in.’ , at solid-melt interface, degrees.
,p
constant defined in E q I , lbf-sec”/in.2 = mass rate of leakage flow per unit length
WE^
constant defined in E q 1, dimensionless. along screw, lb/sec-in.
screw speed, RPM. W8 = melting rate per unit length along screw,
pitch length of extruder screw, in. lb/sec-in.
volumetric flow rate per unit length
along screw: QE8,for leakage flow; Q8, REFERENCES
for melt flowing into the melt pool, 1. E. C. Bernhart, (ed.), “Processing of Thermoplastic
in.’/sec. Materials,” 236, Reinhold Publishing Corp., New York
temperature: T, mixing-cup average in (1959).
2. B. H. Maddock, SPE J., 15,983 (Nov. 1959).
melt film; Tb, of barrel; T,, of solid poly-
3. B. H. Maddock, SPE J., 15,383 (May 1959).
mer feed; T,, of melting (at solid-melt
4. L. F. Street, Plastics Technology, 5, 27 (Aug. 1959).
interface), O F .
5. J. M. McKelvey, “Polymer Processing,” 1st ed., 279, John
constants defined by Eqs 22 and 16: Ui Wiley and Sons, New York (1962).
lbf/sec; U2,dimensionless. 6. E. C. Bernhart and J. M. McKelvey, SPE J., 10, 419
velocity: v b , of barrel relative to screw; (March 1954).
7. Z. Tadmor, Polymer Eng. Sci., 6, 185 (1966).
vj, of shear in melt film; Vj, of shear
next to barrel; V,,, of solid bed along 8. Z. Tadmor, et al., Polymer Eng. Sci., 7, 198 (1967).
9. C. I. Chung, Modern Plastics, 45, 178 (Sept. 1968).
z-axis; V,,., of solid bed along z’-axis,
10. I. Klein and D. I. Marshall, (eds.), “‘Computer Programs
in./sec. for Plastics Engineers,” 1st ed., 254, Reinhold, New York
width: W, of channel at solid-melt in- (1968).
terface along x-axis; Wb8, of channel at 11. Z. Tadmor and I. Klein, Polymer Eng. Sci., 9, 1 (1969).
barrel circumferentially; Who, of bottom 12. D. R. IIinrichs, “A Melting Model for Plastifying Ex-
of channel circumferentially; W,, of truders,” M.Ch.E. Thesis, Univ. of Virginia (1969).
flight land normal to flight; wg, of chan- 13. M. S. Hodgman, (ed.), “Mathematical Tables from
Handbook of Chemistry and Physics,” 10th ed., 297
nel at solid-melt interface circumferen- Chemical Rubber Publishing Co., Cleveland ( 1957).
tially; W,,, of channel at arbitrary point 14. R. B. Bird, W. E. Stewart. and E. N. Liphtfoot.
circumferentially, in. “Transport Phenomena,” 94, John Wiley and So;, New
York (1966).
rectangular coordinate, cross-channel 15. Ibid., 286.
direction, in. 16. D. I. Marshall, Western Electric Co., personal communi-
solid bed width: X , at solid-melt inter- cation.
face along x-axis; &, at solid-melt inter- 17. I. S. Sokolnikoff and R. M. Redheffer, “Mathematics
of Physics and Modern Engineering,” 2nd ed., 657,
face circumferentially; &, average cir- McGraw-Hill, New York ( 1966).
cumferentially; x81, of solid-melt inter- 18. Z. Tadmor, personal communication.

POLYMER ENGINEERING AND SCIENCE, SEPTEMBER, 1970, Vol. 10, No. 5

You might also like