ON ISOTHERMAL FLOW

OF VISCOUS LlQUIDS
THROU GH

SCREW

PUMPS
I

PROEFSCHRIFT
TER VERKRIJGING VAN DE GRAAD VAN
DOCTOR IN DE TECHNISCHE WETENSCHAP
AAN DE TECHNISCHE HOGESCHOOL TE DELFT,
OP GEZAG VAN DE RECTOR MAGNIFICUS
IR. H.

j. DE WIJS, HOOGLERAAR IN DE AFDE-

LING DER MIJNBOUWKUNDE, VOOR EEN

COMMISSIE UIT DE SENAAT TE VERDEDIGEN
OF VRIJDAG 11 SEPTEMBER 1964 DES
NAMIDDAGS TE 2 UUR

DOOR

MAX LORENZ BOOY

WERKTUIGKUNDIG INGENIEUR

GEBOREN TE ROTTERDAM

DIT PROEFSCRIFT IS GOEDGEKEURD DOOR DE PROMOTOR

Prof. Ir. J. O. HINZE

On Isother mal Flow of Visc ous Liguids through Screw Pumps

CON TEN T S
Chapter 1.

Introduction to Screw Pump Theory

1.1

Introdu ction
Screw Pump Geometry
1.3 Screw Pump Theory for Prismatic Rectangular Channels
1.3.1 Assumptions for Re ct a ngul ar Channel Theory
1.3.2 Fr ame of Reference
1.3.3 Development of Rectangular Channel
1.3.4 Equations of Motion
1.3.5 Flow through Channel with Rectangular Cross Section
1.3.6 Separ a ti on int o Si mpI er Solutions
1.3 . 7 Si mplified Fl a t Pl a te Theory
1.3.B Shape Factors of Re ctangular Channel Theory
1.3.9 Pressure Distributi on of Simplified Theory
1.3. 10 Dissi pated Energy of Si mplified Theory
1.4 Other Theories
1.5 Scope of Investigation
1.2

Chapter 2.

Intr oduction
Assumptions a nd Frame of Reference
2.3 Equations of Motion
2.4 Equat i ons of Motion for Large Aspect Ratios
2.5 Velocity Di s tributions
2.6 Scr ew Pump Equation
2.7 Curvature Factors
2.B Influence of Side Wall s
2.9 Pr essure Di stribution in Helical Channels
with l ar ge Aspe ct Rati os
2.10 Energy dissipated in Channels with large
Aspect Ratios
2.11 Corrections for Finite Land Width and
Multiple Flights

3.1
3.2

3.3
3.4
3.5
3.6

1.2

1.6
1.7
1.7
1.B
1.9

1.13

1.14

1 . 16
101B
1.20
1.23
1.29
1 . 30

Influence of Channel Curvature on

Performance of Screw Pumps

2.1
2.2

Chapter 3.

1.1

2.1
2.2

2~4

2.10

2.12
2.13
2.19
2.23

2. 26
2.30

2.33

Leakage Fl ow in Screw Pumps

Introduction
Defi ni tion of Leakage Flow
Pressure Distributi on with
Leakage Fl ow Rate
Leakage Fl ow Rate and Heat
Leakage Layer on inside of

Rate
Leakage Flow .
Transfer
Barrel

3.1
3.2
3. 3
3.7

3. 1 5

3.1B

Contents (continued)
Chapter 4.

End Eff ects in Channels

with Large Aspect Ratios

4. 1 . Intrduct ion
Geometry of Screw Pump Channel
4.2
Average Velocities and Velo city Potential
4.3
4.4 Superposition of SimpIer Cases
Symmetry Considerations
4.5
End Correction Factors
4. 6
Nunieri cal Method
4.7
4.8 Functional Relation between FDE ahd FpE
4.9 Calculation Results
4.10 End Effect for Large Channel Ratios
4.11 Pressure Distribution and Streamlines
Chapter 5.
5.1

5.2
5.3
5.4
5.5
5.6

. 6.1
6.2
6.3
6.4
6.5
6.6
6. 7
6. 8
6. 9

4.1
4.2
4.4
4.8
4.11
4.14
4.16
4.18
4.23
4.30
4.37

Experimental Verification of End Effect

Introduction
Model Channel
Simulation of Screw Pump Channel
Streamlines
Pressure Distribution
Experimental End Effect Factor

Chapter 6.

Page

5.1
5.1

5.2
5.4
5.5
5.6

Numerical Solution of
Partial Differential Eguations

Introduction
Principles of Method
Extension to Families of Regions
Dimenmonal Analysis and Shape Factors
Accuracy of Numerical Method
Sha~ Factors for Tubes with
Rectangular Cross Sections
Errors and Precision
Calculation of End Effect Factors
Additional Examples

6.1
6.3
6.10
6.12
6.16
6.17
6.21
6.34
6.43

Surnmary

I-IV

Samenvatting

V-IX

Nomenclature

X-XII

References

XIII-XIV

1.1
CHAPTER I
3

1.
1.1

Introduction to Screw Pump Theory

Introduction

Screw pumps have become increasingly more important with the

growth of the plastics and synthetic fibers industries.

Screw

pumps are used almost exclusively to pump liquids with very

large viscosities - in the liquid-filled dis charge zone of a
plasticating extruder, for example.

Polymeric materials are

melted in plasticating extruders and discharged at high pressure

througp a die to form the desired shape.
This Thesis is concerned with the isothermal flow of very
viscous liquids through screw pumps.

Present theories are based

.~

on a number of simplifying assumptions or restrictions.

A more

general theory should correctly account for the true geometry of

the channels, the physical properties of the liquid, the temperature changes in the liquid, and the operating conditions.

For

the present, such a general theory cannot be given.

For instance, in present theories the helical channel is almost
always replaced by a straight prismatic channel with a rectangular cross section.

The assumed inlet and dis charge channel

geometry differs in these theories from the actual geometry.

The simplified geometry can lead to discrepancies between
calculated and observed results.
The non-Newtonian properties of the liquids tend to be blamed
for some of these differences.

Such reasoning is acceptable

only when the flow analysis accounts for the correct geometry
of the screw pump channels.

1.2
One purpose of this Thesis is to provide a more complete theory
through the elimination of some of the simplifying assumptions
about the screw pump channel geometry.

The influence of the

differences in channel geometry on screw pump performance can

best be studied when other factors are constant.

These influ-

ences will therefore be investigated for isothermal flow of a

Newtonian liquide
Flow through a helical channel bounded by cylindrical surfaces
and by helical flights will be discussed in Chapter 2.

The

influence of the obliqu inlet and dis charge ends of the channel
on screw pump performance is given in Chapter 4.
A comparison between these new theories and the current theory
requires a fuller discussion of some elements of the screw pump
theory based on isothermal flow through rectangular channels.
Discussion of the existing theory is preceded here by a
description of the geometry of screw pumps.
1.2

Screw Pump Geometry

Screw pumps generally consist of a single screw rotating in a

cylindrical barrel.

Twin parallel screws are sometimes used

in plasticating extruders but never in screw pumps.

A schematic diagram of a screw pump is shown in Fig. 1.1.

~FEEO 90QT

~ CRw \(00'I

SvRrI\C~

RAD'AL C\.SAl?A..",Cf. - C

~I(?,.

1.2-

T\.iP\CA\...
W\14

2.

a.~Rf(t.\..
PC\R~L\.E\..

A...,t:>

sceEW

f=L.\~~T~

1.4
A positive inlet pressure is used to insure that the inlet zone
of the pump is completely fil1ed.

Th screw fits with a small

clearance c in the cylindrical barrel.

Rotation of the screw

causes the liquid to move to the discharge port.

Figure 1.2 shows a cross section of a typical barrel and screw.
The screw has an axial flight land width e.
Screws can have more than one parallel flight, as this i11ustration shows.

Two parallel channels are formed between the

barrel, the screw root surface, and the flights.

The channel

depth h is the radial distance between the barrel and the screw
root.

Both channels are machined with a lead t.

The pitch then

equals tin for a screw with n parallel flights.

The clearance is generally very small compared to the barrel
diameter D.

The helix angle

te 0

at the flight tip is for a

negligible clearance given by

<Po

= t~1 (t/1I D)

The helix angle of the screw flights changes along the depth of
the channel.

The helix angle at a radius r is

'e~. +~I(t)2)h).~' [~-t.~~oJ

The local helix angle 'Pr increases \'lhen r decreases.

The channel

width bo)developed at the flight tips is, for a negligible

clearance, given by

1.5
es
A simi lar chan nel widt h, deve loped at radiu s r, becom

(1.4 )
is defin ed
The effe ctive , or fligh ted, RxiR l leng th of the screw
as L. The leng th of a fligh t tip is then

Lc.o ~

(I. s)

barr el
This leng th is also the leng th of the chan nel at the
r becom es
dinm eter. The chan nel leng th deve loped at a radiu s

root, since
The chan nel leng th then decre Rses towa rds the screw
sin~r incre Rses with a decre ase of r.
screw root
A fille t radiu s is alwa ys used at the tran sitio n from
t is
to screw fligh t. The fille t radiu s of the trail ing fligh
not nlwa ys th same

dS

that of the leadi ng fligh t.

s have
numoer of othe r screw geom etrie s are poss ible. Screw
the screw .
been made with a lead that varie s along the leng th of

J\

These
The depth h v:1ri es in some screw s along its leng th.
s of
spec ial geom etrie s, gene rally restr icted to melt ing zone
extru ders , will not be cons idere d in this anal ysis.
only one
The anal ysis will also be restr icted to screw s with
will be
fligh t. Corr ectio ns for addi tiona l para llel fligh ts
r,iven wher ever nece ssary .

1.6

LJ

3crew Pump Theory for Prismatic llectangulnr Channels

Isothermal flow of very viscous liquids through screw pumps and

metering zones of plasticating extruders has been studied by
v.lrious investigators, a few of whom are listed in the references (1, 2, 3, 4, 7).

A screw pump equation which relates the

flow rate and the generated presstire to speed, screw pump

geornetry and physical properties of the liquid is derived in
erl ch theory.
In the

exi~g

theory the heli cal screw channel is replaced by

n straight prismatic channel with a rectangular cross section.

A very large channel aspect

ratio : o~ la~ge

width todepth ratio

of the rectangulnr cross section, is assumed in the simplest

theory.

Flow rates for that case can be based on well known

velocity distributions for flow ' between parallel plates.

The theory based on flow bet,."een pnrallel plates (I) will
further be referred to as the simplified or flat plate theQry.
lt leads to a relatively simple screw pump equation which is
very useful as a reference equation.

The results of more

refined theories are usually compared to the screw pump equation

of the simplified theory by the introduction of correction
f.J ctors.
The rectangular channel theory, which for very large aspect
rntios deGenerates to the simplified theory, is first discussed.
Thnt discussion is not restricted to the published literature.
The screw pump equations will be derived and terminology comrnon
to nll theories will be introduced.

1.7
1.3.1
~he

Assumptions for Rectangular Channel Theory

following restrictions are imposed in the isothermal rec-

tangular channel theory:

a.

The viscosity of the liquid is uniform and not influenced by

shear r ates.

b.

Inertia and body forces are neglected.

c.

The liquid is incompressible.

d.

Velocity distributions in channel cross sections with planes

perpendicular to the centerline are congruent.

e.

The helical channel can be replaced bya straight prismatic

channel with a rectangular cross section.

One side of that

rectangle is equal to the developed width b o (1.3), the

other is equal to the channel depth h.
f.

The channel has no rounded corners.

g.

End effects are neglected.

For now it is further assumed that the leakage flow rate through
the flight clearances ca n be neglected.

This leakage flow rate

will be treat ed separately in Chapter 3.

1.3.2

Frame of Reference

The velo city distributions in the channel can be described

relative to any desired frame of reference.

The analysis is

simplified when all velocities are described relative to the

screw.

Such a velocity distribution is the one seen by an

observer who rotates with the s crewo

The' screw appears to him

to be at standstill, while the barrel and the gravity field seem

to rotate in the opposite direction.

The velocities relative to the screw are identical, when either

the barrel is at standstill and the screw rotates, or the screw
is at standstill and the barrel "rotates with an equal speed in
the opposite direction, provided inertial and body forcs can be
i gnored .

This is precisely the second assumption made - tht

Reynolds' number must be small and viscous

fo~ces,

compared with

gravity forces, large.

Velocity distributions will always be described relative to the
screw.
1.3.3

Developmen~

of Rectangular Chahnel

"The liquid occupies an annular space between two coaxial cylinders, except for the space occupied by the flights.

For large

channel aspect ratios this annular space is very similarto the

space between two parallel plates.- This' similarity suggested
development of both barrel and screw root stirfaoe into flat
plates to early investigators.
The development of the cylinder with diameter D, which coritains
the flight tips, is shown in Fig. 1.3.
" In this theory it is assumed that the identical development
pplies to the flights at

th~

root diameter, D-2h.

Ah error is

introduced here since a cylinder with a smaller diameter is

actually developed. "The error increases with the depth of the
channel

1.9

I~---+---

lID

~---'-~---~ T D '"

,..,.
L

FIG. t. 3 _

DE-VE.lOPfV\Et-.J

fo

SIiJGl.~

.-

OF

SCf<E.W FlIGl-lrs

FLIGI-\"TE D

SCR.~vJ

1.10
The flights have a uniform helix angle in this development" which
is equal to the true helix angle at the flight tips.

The

channel width perpendicular to the flights is nowindependent ot

the depth in the channel and equal to the width b o (1.)

at the

periphery.
The screw channel can also be figuratively unwrapped from the
screw.

A perspective projection of the straight prismatic

channel is shown in Fig. 1.4.

The barrel, which now becomes a

flat plate, moves with a uniform velocity1f DN over the flight

tips.

The direction of that velocity makes an angle

the flights.

y, 0

with

The barrel velocity can be resolved into com-

ponents Uo and Vo' respectively parallel with and perpendicular

to the flights.
1.3.4

Eguations of Motion

A Cartesian coordinate system (x, y, z) is introduced in

Fig. 1.4.

Velocity components in the x, y, z directions are

respectively u, v, and w.

The velocities at z = h must be

equal to the barrel velocities and must vanish at all other

channel walIs.

The velocity distributions must satisfy a speci-

fied pressure difference between inlet nd discharge.

In the absence of body and inertia forces, flow of an inoompressible liquid is governed by the Navier-Stokes equations

l.ll

".

1 . 12

op

'X.

JA-

2-

\l u.,

( 1.,0..)

'2.

jJ'- Vu

0'-]

~ -: : f-"L

(1.7 b)

vu.)
'L

(1./ c )

0"2

where

\J

2-

-o'l...

d7..
vX?.

'";)'1'2..

Cl?...

-+

o ~'2..

' One of the assumptions states that velocity distributins in

planes perpendicular to the centerline of the screw are congruent .

The velocities are then the same at all points On the

l i ne TTI parallel -to the x-axis and are independent of

=.

x.

Thus

- - ----- _0

ox'1. -

Fr om (1.7 )' and (1. 8) ,

l ( ~)
=
l..-\.
ox
oX
JA V ()()x
2

== 0

The deri'Vativ dp/<Jx is then independent of x, y, and zand

must be a constant.

Substitution of (1-.8) in (1.7) reduces

the Navier-Stokes equations to

(1.100-.)
(I ~ 101,)

(I.IOC)

The distance ' between points Tand Tl respectivelY ,at inlet and
"

discharge {Fig. , lo4} is for Y and z = constant' always ,equal to

the developed channel length Lca (105).

bet ween such points is the same for any

dpfo x

The pressure difference

va~ue

of y or z since

constant 0 In the rectangular channel theory that .

difference is equated to the pressure difference

the screw pumpo

Po

generated by

We will show later {10309} that this theory

does not predict uniform pressures along the inlet and discharge
ends o The effect of the ends has here been neglectedo
The uniform pressure gradient in the channel directionbecomes
with {lo5}

(1.11)

where Aa is introduced to cancelr- in (1010a) 0

1.305
The

Flow Through Channel with Rectangular Cross Section

flo~

rate through the screw pump is the flow rate relative

to the screw in the x=directiono

This flow rate Q is equal to

1.14
The partial differential equation for u (1.10a) is independent
of x, since 0 ui Q x = constant.
independent of x.

The boundary condi tlons are also

The u component is then a fnction, u(y,z),

that is completely determined by this differential equatin and

its boundary conditions

\A.( ~) 0) = LA. ( 0. 2)
u.( ~ I h ) = lto

==

lA (

1 0 , ) =

(I. \3)

(I. 14 )

The remaining differential equations need not be solved, if one

is only interested in the flow rate.

The problem is then

reduced to the two-dimensional problem of dterminihg the solution

to

l'u.
"2_+OIA
.:
~1..

0 ~1.

(I. ,s) ,

for the boundary conditions (1.13, 1.14).

1.3.6

Separation Jnto SimpIer Solutions

The problem can be simplified by rsolving (1.15) into two simpler

equations.

Consider determination of the solution to

(LI~ )

vr'2. uo -= Ao
for a set of boundary conditions, here symbolized by BG o '

Assume now that the solutions to m-different Poisson equations

of the form ,

with associated boundary conditions BCi and associated constants

Ai (i = 1, m) can be determined.
of m-solutions.

Let u.~ (i = 1, ro)

b~

the set

The desired solution Uo is thert

""""

uo=~Lt,L.
(.;= \

CL l'b)

. 1015
~

provi ded that

Ac=

L.
(.-=-1

A.'-

Cl. \'\ )

""'-

Beo

and also

~ Be-\,...

U 2o )

~,

Thi s can be proved by direct substitutiono

F'o r our purpose , the problem is separated then into finding the
solution to these two partial differential equations and associ =
ated boundary conditions
lA,(o) 2'):. 0
U,(bol~ ) = 0

\AI(~IO)'='O

'~ , l~th\=l1o
The required solution is

Uo =

uI + u20

The separate problems have t he fo11owing simp1e physica1

ing o

In the first

problem~

me~n~

the barrel z = h i s dragged over the

channel with a velocity Uo in the absence of a pressure gradi ent o

For this case, the integral (1012 ) of the velocities is called
the drag flow r ate, Qn 0
The se cond problem des cribes the familiar case of a velocity
distribution in a conduit with a rectangular cross section and
-. '::

~:--stationary

walls caused by a uniform pressure gradient in the

channel directiono

Veloeities are negative for a positive

pressure gradiento

The pressure back flow rate Qp is now

commonly defined as the flow rate caused by a negative pressure

gradient , i 0e ~, Q-p ') 0 when in opposi te direction to

and

Qn

1.16
The total flow rate, neglecting the leakage flow rate, is with
this definition

Both differential equations (1.21, 1.22) can be solved by the

method of separation of variables.
analog in heat transfer (9).

The first problm has an

The second

AiryT s stress function distribution

rectangular cross section (10).

~n

pro~lem

has an analog in

a twisted bar with

The resulting equations will

be shown af ter a discussion of the simplified theory for flow

through rectangular channels with large aspect ratios.
1.3.7

Simplified Flat Plate Theory

A very large aspect ratio b/h, which approaches an infinite

ratio, is assumed in the simplified theory.

Later it will b

shown that there is only a small difference between the solution

for large aspect ratios (bo/h> 10) and that for an infinite
aspect ratio.

The theory th en applies to large aspect ratis.

The differential equations (1.10) can be simplified when aspect

ratios are large.

The channel then closely resembles the space

between parallel plates.

For that case, all velocities are

parallel to the barrel, except in small regions near th

flights.

The width of these regions is of the order of h.

The

influence of these regions becomes insignificnt f ot' very large

aspect ratios.

The velo city component w, therefore, vanishes for

a sufficiently large aspect ratio in ~ll but an insigniti~ant

region.

1.17
Further, the u component 15 zero at the root z

0 and equal to

Uo a t the barrel z = h~ The der1vat1ve 0 u/~y 15 1ns1gnif1cant

compared with ujd z. The pressure gradient p~ z van1shes for

w=

o.

The parti al d1fferent1 a l equat10ns (1. 10 ) th en reduce to

ol2~
-

t)

~
lA"b)( -- /-

ct ~ ..

;::

rAA.- Aa
(1:2Sb)

( I. '2.""c ')
Here the partial velo city derivat1ves were replaced by regular
derivat1ves.

The pressure 15 not a function of z, since op/~ z

van1shes, while

opf;:; x is constant.

15 a funct10n of z only.

The r ight-hand s1de of (1. 25b)

That equat10n can be sat1 sf1ed onl y

when pjo y 15 also a consta nt

Both equat10ns can then be

solved by s1mple 1ntegr a t1on.

The solut1on to equat10n (1.25 a ), 1n analogy to t he twod1mens1onal ca se, can be forma lly regarded as the superpos1tion
of a dr ag flow and a pressure back flow d1str1but1on.
follow1ng equat10ns must then be solved.

Integr at10n y1elds

The 11near d1str1but1on UI yields the dr ag flow

Qt>o =

h b o Uo

The

1.18
The parabolic distribution u2 results in a pressure back flow

",

~?o=

Po Sv....,~o

h3 bo
17..

.,AA L

Substitution of

b o :: (t _ e.)
Uo = 1T!) t--J
t

:: 11 D

c.,o~ '10
Co S

Y,o

-t~teo

in the equations for QDO and QpO results in

2= .1t1..
'2 L,.,...lfo ~s~. (\ - e/t.) h b ~
1
Qpo =- 1& sVvt~o (, - e/t) h j)

QDO

:QL

The screw pump equation for the simplified theory without leakage
flow is now
'2..

YSv...~o COS'fo(I - e/t )h])21\) _ ~ S;.tl'o (I-e/t) h'!) ;';..

Go< =

This is the reference equation with which results of btner

theories are compared.
1.3.8

Shape Factors of Rectangular Channel

Theo~y

The partial differential equations (1.2l, 1.22) for a rectangular channel yield a drag flow rate QD and a pressure back flow
rate Qp that can be expressed in the corresponding flow rates of
the simplified theory through the introduction of correction
factors FD and Fp.

Then

For that case, these correction factors are called the shape
factors.

The shape factors (4) are the following functions of

the aspect ratio

----

---------------------------------------------------------~

/u

----L--, .~

_D

~I

1.19

Thenc shn:pe fnctors are shown in fir;. 1. 5 as function of the

reciprocal aspect ratio h/hO. They reduce for large aspect
ratios to

Ft) =

I.

Fp ::

'.0 -

0 -

o. . ( e../ b
0.(,'1

0 \

(h/ho)

1.0

o,~

o.

.L------r------~--~--~+_------~------~

o.~

o.

------+-------~------_4--------~----~~

h = I ,
b o ASPtC.1"' Rld',Q

~-

0.4

0.'2.

0.4

o. "

FH. 1.5 _ SHAPE: F.l\CTQQS

o_~

\.0

1.20

The3e equations, nnd Fig. 1.5, show that the equations of the
rectnnr,ulrlr channel theory approach those of the simplifietl
theory for larger aspect ratios.
~2

Pressure Distribution of Simplifid Theory

It Wl1e ahown betore that the gradienta':>p/l:Ix An OpAy ar. 'both

constant in the simplified theory, while (:, pfo z vanishes.

The

pressure distribution in the channel is then a linear function

An expression will now be deri ved for '0 pfo y.

of x and y.

The transverse velocity distribution can be obtained by integration of (1.25b) for the boundary conditions

\.r(

~ I ~)

( 'dl h)

= \10

This distribution becomes

The transverse flow rate per unit flight length is then

h

q ~ (' \Tcf"
(l

Jo

= J..
2.

hU. _ h'l.> ~
0

(\.4'2.)

1'ljV'- >'}

This flow rate is independent of y oand applies equally weIl to

the flight face y = O.

This flow rate must be zero when no

leakage is considered, since there cannot be oa flow rate crossing the flight.

Then

~ = ~\Yo
'0'1

h~

The pressure distribution is given by

p= x~

+'d1;;

+ PI

1.21

where Pl is tpe pressure at x = 0, y =0.

isobar is constant.

The pressure along an

The equation for an isobar is then given by

QE dx '+

eX

The slope .o f t1)e isobars is f r om {1 . 11, 1.43}

~olx

()pl~'x

Po h2-

= - ----"---c,11';U DI'JL

o~/v"d

The' isobars of the simplified theory .are parallel lines with a

s"lope that

is

f ul.l y det ermi ned by the geometry of the pump and

the operat ing condit i onc

We now 'introduce the back flow coefficient ~, defined as the
ratio of the back flow QpO to the drag flow QD00
(1032, l '. J ;, )

It is from

'2..

Pa h i~lfo
Gl1)..A DN L

This coefficient characterizes the mode of operation of the screw

pump.

It is zero for pure drag flow and becomes 1.0 when the

discharge is shut .

The drag flow is equal to the pressure back

flow when the pump operates ' against the shut-off pressure PS.
The pressure back flow coefficient is then also

It can be regarded as a dimensionless pressure equal to the ratio

of the operating pressure Po to the theoretical

shut~off

pressure

Ps that could be obtained at the same speed if the viscosity

could be kept constant .. .
Comparison of {1.46a } to
bars is also

(1~45)

shows that the slope of the iso-

~
= -_:--

tCNY\..\eo

~.

1.22

The slope of the isobars is constructed in Fig. 1.6 for a series

of~ values tor a screw pump with a helix angle of 17.65

(t/D = 1).

,.1.0

-\.0

SL.OPES o:- . l

~
Fol2. C\!='FE-i2EIoJT

:tso E.C2$

VAI...vE-s or- ~

FL,\G 1otT'

FI< ..

\.Co _

FvNCT I O~

SLoPE-S ot: \SOSAI2S

o: f'YIoofo

AC;

~ Of'E~TI O~ :-Q~ .~=I7.bSo

This .figure shows ho~ the slopes of the isobars change with
operating 'conditions.

.,

The slopeJl.4.7) corresponds to the slope

of the ends, tan (p 0' only when

'-'

= -

i (" .

-LCXM. "t'o

1.2)

Such an operating condition demands ~ = -0.1015 for ~o

17.65, or a discharge pressure that is lower than the in1et

pressure.

The slope of the isobars can, for norma1

operati~g

conditions (~>0.1), never match the slope of the channe1 ends.

In actual operation, one must expect constant pressures at the
channe1 in1et and discharge ends, when the connections to the
screw pump housing do not restrict flow.

The simp1ified theory,

and for that matter the rectangular channel theory, do not

satisfy these boundary conditions.

An analysis is given in

Chapter 4 for channels with large aspect ratios for flow with
uniform pressures along inlet and discharge ends.
1.).10

Dissipated Energy of Simplified Theory

In the previous analysis, it was assumed that the temperature of

the liquid does not change along the length of the screw.

Very

viscous liquids are sheared and one must expect a temperature

increase, unless the pump is cooled.

The assumption that the

pump operates isothermally is generally incorrect.

It is

important to make an order of magnitude estimate for the

expected temperature increaGe.
In the fol1owing, we will discuss the average temperature rise
for a uniform viscosity liquid that is pumped without heat
exchange to the surroundings.

Leakage flow is again neg1ected.

1.24
The energy dissipated in the liquid can be conveniently cal culated from the dissipation function which, for the remaining shear
rrltes

) uI

Vz and'O vI z, is given by

~ =? [\~~)"-+ n~n .

(1.49)

The energy per unit time dissipated in the channe1 is

EI

fJ~JLco
q;
o

ol)(

d'd

~-:r-

(1.50)

The dissipated energy is for the same flow rate and the same
eenerated pressure influenced by the magnitude of the back
flow coefficient ~.
e1iminate

Po

To show this we first use (1.46a) to

from (1.28).

The velocity. distribution

be~omes

(1.51a)

The distribution of the v-component, given by (1.41) and

(1.43), reads
(1.51b)
The shear rates are then
-

~o [I - ~~ (1- 2 ~

lJ

(1.52)

'2*[!'(~)-I]

(1.53)

Substitution in (1.50), integration and subsequent substitution

of equations for bo, Uo' Vo' and t yields
E \ ::: 11~;tA- 1)'31-.\2.hL (I - ~/~)__ [(I t ~ 01. ) fJ)~'1..'(o

.
4-

4 s ~\:, ]

(1.54)

The energy Ep' per unit time availab1e at the discharge,is now
(1.55)

1.25
Substitution of ' (l.32) and elimination of Po with (l.46a) yields
3
3 ~
.
\ '2..
Ep: 3TT;Vv'D hN L(\_e.li) ~ ( 1_~)CoSl.fo
Comparison of (1 . 54) and (1 . 56) shows that El can be written as

Write

_ . (I + ~ ~2.) (.oS 2. ~ a +4 si.M,'1.'Po...

k
o

:s ~ (, _~ ) c.os,'1-XOo

Then
. The energy E.p is also the energy dissipated in a pipe or other
restriction, when a flow rate Q requires a pressure drop
between inlet and discharge .

Po

The ave rage temperature increase .

for flow through a pi pe can be calculated from the heat balance

, or

: This average temperature . increase is of the order of 2-3C . for

polyrners for PO= 1 , 000 psi ( 7 010 6 N/m2) .

The ave rage tempera-

ture increase in the screw pump from the energy dissipated in

the ch. nnel is Ko t i mes that temperature increase o
Additional energy is

~enera

ed On t he fl ight clearances .

The

velocity di stributi on i n the cl earances is again the superposition of a drag f l ow a nd a pre s sure back flow distrj.bution .
This back flow is generally small compared with the drag flow,

rp..,

v-

1.26
as will be se en in Chapter 30

This energy can then be based on

the drag flow shear rate along, which is

(I.GI)
Shear rates in the clearances are at least one magnitude larger
than shear rates in the channel.

A non-Newtonian liquid will

have different viscosities in clearances and channels.

Here we deviate rrom the original assumptions and introduce a
viscosity ~F for the liquid contained in the clearances.

The

volume in the clearances is

LZ

= TI D L c (e /t )

The energy per unit time dissipated in the clearances is

o

The

1~F D3

1\,)'4

(e/t\

EF = \(jAF Or = - - - - C - - - relative importance or this energy is shown by

the ratio

+- 3 ~'2.) Cos.'-<Po + 4 s~'2.~~

For example, this ratio is
0.1, ~

0,

Cfo =

0 008 for

;V--F/)A-

c/h = e/t =

17 065 0 It can assume larger or smaller values o

The ave rage temperature increase from El and EF becomes

The energy coefficient KO {1058} is a runction of ~ and

Fig. 1.7 shows that coefficient for different t/D ratios .

~o o

The

energy is then a minimum for back flow coefficients of the order

of 0.3.

The change of KO is small near that minimum.

tive ~ corresponds to a negative pressure PO .

Ka is negative for that caseo

A nega-

The parameter

The dissipated energy is then

positive, as it should be, while Ep is negative.

1.27

B.o

~.o~------~----~~--------~------~---j

0.'2.

F\G.

(J.S
1.0

It)~ 2.0

\.7 -

-4.0

1----+-----t---t--t-1"r---1 - (:,. 0

O.~

ENE;RG'1 Coi.~r:\C\E..J K. o

A ~ rUN C T\OJoJ O~

LS

I---+-----t"\--t-t-r--

0.+

e,ACA<. ~\.ow coF. ~

1.28

o.~o ~------~~-------4--~-----+--------+----

----~

0.2.0

I~--l--I---++----+---~+-~r\

---t-----;

O.\S ~~~--~-------+------~~~~~rt
------

--1

0.\0

o.o~ ~~-----+---------+---------r--------

-r--~~~

0.2.
F\G. I. ~ _ ~"1l) I2AUL..\C ~

0,6

CI~""C.~ O~
t\-.J'f;<'1 t> \'S.S\P Af C

Ff I

SCR~WPUMP ) N(.GLE c..TII.J Gr

\\..J F\..IC ,HT C\..~~~A.l.JctS.

o,ca

\.0

1.29

An order

o~

magnitude estimate of the average temperature

~rom

increase can be made

Fig. 1.7 together with (1.63), when

PO' ~ and the parameters in (1.63) are known.

This increase

is of the order of 8-10 C . per 1,000 psi (per 7.10 6 N/ m2)

The e~ficiency of the screw pump becomes for ~ >0

"Y)

Ep

Ep

1-

E, ,... EF

I +- K

This ef~iciency i s shown in Fig . 1.8 ~orEF =

and no leakage

~low.

is only slightly

The position

o~

the maximum

dependent on the helix angle.

e~~iciency

This maximum occurs at values of

~ between 0.33 and 0.40.

Note that the pump be comes a prime mover, when both f->
K

>- 1.

<

0 and
I

The excess energy is in that case

EfV\

.=

- , ( \ 4-

1<)

(Q

Po

The efficiency as a prime mover becomes ~or ~ F

1+ k
The energy equations can readily be converted into equations
~or

1.4

calculation

o~

the drive horsepower.

Other Theories

A number

o~

other theories have appeared .in the literature.

McKelvey (6) discussed adiabatic operation

de~ined

o~

screw pumps,

as operation without heat exchange with the surroundings.

Performance is then

in~luenced

by the temperature increase and

associated viscosity reduction.

ColweIl and Nicholls (14) analyzed

~low

through screw pumps with

large aspect ratios when screw and barrel

sur~aces

at arbitrary temperatures andthe viscosity is a

are cntrolled

....'.

.! 1.30 "
function oftemperatuI"e and shear stress.

They considered shear

stresses in the channel directiononly.

De Haven (13) derived

equations for i "sothermal flow of a pseudoplastic liquid through .

an extruder.
,

A number of theories on flow of non-Newtonian liquids (13, 14,

15, 16) have been published.

These theoriei have in common that

shear rates in the channel direction only areconsidered.

Such

a one-dimensional model is incorrect, since transverse shear

rates are ." significant.

Consider, for instanee, the shear rates

l

for a Newtonian liquid, given in equations (1.52, 1.53).

transverse shear rate at the barrel is larger for

t/D

~ =

The

0 and

1.0 than the shear rate in the channel direction by a

factor 1.30.

A substantial error is introduced when transverse

shear rates are ignored.

1.5

Scope of Investigation

The existing theory, discussed in this chapter, is based on a

channel geometry that differs from the actual geometry.

An

improved theory that accounts for some of these geometrie

..

differene,s will be gi ven in. the following chapters. ' The

~"

influence of these differences is drived for "isothermal flow of

Newtonian liquids.

Four distinct problems are treated in this

Thesis.
The first problem deals with the influence of the
the walls of ,s crew pump channels.

In existing

c~rvature

theor;i.~s',

the

~elical channel is replaced by a straight prismatic channel

of

f.-3t'
with a rectangular cross-section . Flw,-through channels bounded

by cylindrical barrel and srew root surfaes and by two heli cal
flight surfaces :is discussed in Chapter 2.
The second problem is oncerned with leakage flow through flight
The existing theory is extended in Chapter 3.

clearances.

Leakage is responsible for the formati.on of a thin liquid layer,

that remains near the barrel surface.

The layer thickness, which

must influence heat transfer to the liquid, is shown to be a

function of the geometry of the screw, of a viscosity ratio,
a,nd of the mode of operation of the pump .
The third problem deals wi t h the effect of the oblique channel
ends on the performance of screw pumps .

Existing theories do

not yield a channel pressure distribution that satisfies the

pressure boundary condit i ons along the oblique channel ends.
Flow through channels with large aspect ratios i s dlks.cuss'ed in
Chapter

J';'~r

for the proper boundary conditions.

Comparison with

flow rates of the simplifi ed theory leads to the introduction of

end effect correction factors .

Pressure distributions and end

effect factors for pressure ba ck flow were experimentally

investigated .

Test results are discussed in Chapter 5.

The fourth problem deals with the numerical solution of linear

partial differential equat i ons .

Calculation of end effect factors

requires the solution of Lapl;c'e's

regions.

equation in a large number of

A computer techni que was developed for the rapid

calculation of linear partial differential equations in large

families of regions.

That method is discussed in Chapter 6.

2.1
CHAPTER 2
2.

Inf1uence of Ohanne1 Curvature ort Performance of Screwpumps

2.1

Introduction

In existing theories, the heli cal channel is replaced by a

straight prismatic channel with a rectangular cross section.
Th1s subst1tute channel resembles the actual channel reasonably

weIl, when channels have large aspect ratios.

width

~o,

channel length Lco and helix

The channel

angle~oof

the prismatic

channel are all uniform and equal to the width, length, and
helix angle at the barrel diameter.

These dimensions are

quite different at the root diameter of the actual channel,

particularly when channels are deep.

The correction factors

Fn and Fp of the rectangular channel theory account for the

finite aspect ratio, but they do not account for the changes
in width, length, and helix angle with channel depth.
In this chapter, flow through helical channels will be discussed.
This analysis differs from previous theories by Squires (17)
and McKelvey (18).

Some of the results given here are the

same as those previously given by the

author (5).

A solution was obtained for flow through helical channels with

large aspect ratios only.

That solution differs from the

solution predicted by the simplified flat plate theory.

The simplified flat plate theory predicts an equation for
the flow rate which is incorrect when used for a finite aspect
ratio.

The shape factors Fn and F p are in the rectangu1ar

channel theory introduced to correct for the reduced ve10cities

near the short sides of the channel.

2.2
The equations, based on flow through helical channels with .
large aspect ratios, contain a similar error, when applied
to channels with large aspect ratios.

Another set of correc-

tion factors FDl nd Fpl would result i-f exact flow rate
equations for helical channels witn finite aspect ratios
",r vjf!

could. be compared to equations for flow through heli cal

channels with large aspect ratios.

These exact factors cannot

be given, since the exact solution was not obtained.

The

assumption is then made, that FD1 = FD and Fpl = Fp Such

an approximation is reasonable for small h/~o values. No
upper bounds can be given for the h/bo ratios for which this
approximation is valid.
2.2

Assumptions and frame of reference

In this chapter,now ofa very viscous liquid through helical

channels is investigated for the following as sumptions:
a.

The viscosity of the liquid is uniform.

b.

Inertia and body forces can be neglected.

c.

The liquid in incompressible.

d.

Velocity distributions are congruent in channel cross

sections with planes perpendicular to the center line.

e.

The screw channel is bounded by the cylindrical barrel

and screw root surfaces and by two sides of a heli cal
flight.

f.

The channel has no rounded corners.

g.

End effecta are neglected.

r BARREL AND OD-SCREW

~OOTSCREW
DIA. ::1
,,
I

//

'

/
/

'1/

I~"A.,;---II--r'--'-It

1_

,
I

//
,

bh P

1-----+-1

Ro

FLiGHT
PITCH
I"

I.

-I

-I

FIG. 2-1 - SCHEMATIC DIAGRAM OF SCREW PUMP WITH INFINITELY

TH I N FLI GHTS

\.V

2.4
In this analysis, the leakage flow rate through the clearances
between flight tips and barrel is neglected.

The leakage is

treated separately in Chapter 3.

Velocity distributions are again studied relative to the screw o
A schema tic diagram of a section of the screw is shown in
Fig. 2.1.

The analysis is initially made for screw pumps with

one infinitely thin flight.

Corrections Ior multiple flights and

finite . flight land widths are discussed at the end of the chaptero
Cylindrical coordinates r, Q, z are introduced with corresponding
velocity components u, v, and w.

The coordinate system and some

of the nomenclature is shown in Fig. 2.1.

2.3

Eguations of Motion

The velocity distributions and screw pump equations must be

based on Navier-Stokes' equations for flow of an incompressible
These equations are, in cylindrical coordinates

viscous fluid.

and in the absence of body and acceleration forces

~ ~ ~ [~'l.~

[V2-1f'

-L~ = IV'r

oe- .

where

:. JA-

'-

ol{"

01J" ]

(2.I~J

V-

+ 1...
dIA. ]
r .... '"()t::7

(2.. \ 6 )

'('1..

1-

J'

fo~

_I 0 (v..V') +

'(

- ..," W

.'7 W'"

= 07-

The continuity equation

IA

y-?.

0"J...- + I 0'11( ..... o'e'l. + l21,..

'r () V"

(2. IC.)

('-.Id.)

an incompressible fluid is

(2, Ie)

2.5
We assumed congruent velocity distributions in channel cross
sections perpendicular to the screw axis.

That assumption

leads to additional conditions that must be satisfied by the

velocity and pressure derivatives.
Velocities are then constant at all points on a helical
with the same lead as the flights .

cu~ve

Such a locus of correspond-

ing points is formed by the interesection of a cylinder with

radius rand a helical surface with lead t.

The differential

equation for that heli cal curve is

The change of the component u can be expressed for r = constant

as

That change must yanish when the increments dz and de satisfy

the equation for the heli cal curve .

Substitution of (2.2) then

gives the additional condi t i on for the velocity derivative

ollA -=

[~~

+ Ro ~f"

~] d& =0

The term in brackets must vani sh s i nce de =f

o.

(2.,,)

Then

Similar conditions can be deri ved for the derivatives of v and

w.

The congruence of velocity distributions leads for all

velocity components to the condition

2.6
Differentiation of (2.4) yields the additional condition

(2. S")
A similar condition can be derived for the pressure gradients .
All velocity derivatives at corresponding points are constant
when velocity distributions are congruent,

Substitution of

these constants in (2.1) yields for points on such a helical

curve

, 21:

or =

CO~ ~r~IVT

CONSANI

CON S/".t-Jj"

1.e
'()~

(2.")

Two helical curves with the same rare shown in Fig. 2.2.

PD

p~ + (~t\ cl e-

Pe

PB + (~ )B d..&

Now

or

From (2.6)

~)'A
( ~e.

.2.)

\. )e B

Then
The . pressure
. difference between any pair of corresponding points
on helical curves with the same radius r is the same when the
difference z between 'the coordinates of the points is the
same.
Two other curves on one heli cal surface with lead t are shown in Fig. 2.3.

The curve through A has the radius r, that through

F a radius r + dr.

2.7

DEvSLOPE-D
H~L'\ C.AL-

CJUI2 VE.

-----!~-

.ft

l)f..V LOPMEIJ'T O~ CjL.IIVD~

F\ G, 2 2. W\TI-\

-,-WO

l-IE:LICA..L. CVICVS

HE-L\CAlSUI'2FAC

-,'

FIG.

'2:'.!;, _ ' HEL\CAL

LEAD

E<St\Jt>.L o

SV(2~AC.

i~A\

OF

W\T~

FL.\GHTS

2.8

0+(5:
I A.
OOI("

Now

PE -= PB + (~)B
From (2.6)
or

...lv-

eN"

l~)A = \~).s
.: . ~ -PB

PF-PE-

The pressure difference btween any pair of corresponding points

with the same difference

z between the coordinates of that

pair is then the same ort any heli cal curve with lead t.

( ~)
c.I

A\.ONG A
~e L.l CAI- CVfZv~

--

QJIoJSTAt-J

Then

('2.~\)

Th change in pressure for r = constant can be expressed as

dp

= ~ ctS +

4i d

Substitution of (2.2) yields the change in pressure along a

heli cal curve

The condition to be satisfied by the pressures is from (2.8)

and (2.9)

Integration of (2.10) along a helical curve between corresponding poihts at inlet arid discharge yields a constant p!'essure
differenc.

This difference is equated to the difference

Po

generated by the screw pump, in analogy to the method followe"d

in the rectangular channel theory (1.3.4).

2.9

+ 'Or
-eJ- L

Then

- Ro ~~0

or

l-~~J.
L
ur-

Later we show that the resulting pressure distribution does not

give uniform inlet and discharge pressures (section 2.9).

End

effects are in this analysis neglected, as they were in the

simpli fied theory .
Elimi nat i on of the

coordinate with (2 . 4, . 2. 5) and substitutipn

of the condition (2 . 11) for the pressure derivatives reduces the

Navi er- Stokes equations for flow with congruent velocity distribut i ons to

where

Ger) = I + Ro2.+~~Q I

rCr)::

2. R0

iewv" \00

\,"4-

r~

The continuity equation becomes with (2 . 1e)

These equat i ons describe the velocity and pressure distributions

i n a plane 8

constant as function of rand z.

2.10
A typical cross section through one of these planes is bounded
by the lines
r = RO (barrel surface)
r = Rl (screw root surface)
z

(flight surface)

z =t

(flight surface)

Boundary conqitions are for that cross section

2.4

u(Rl,z) = 0,

v(Rl,z) = 0,

w(Rl,z) = 0

u(r, 0) = 0,

v(r, 0) = 0,

w(r, 0) = 0

u(r, t) = 0,

v(r, t) = 0,

w(r, t) = 0

u(RO'z) = 0,

v(RO'z) =TrDN,

w(RO'z) = 0

Eguations of Motion for Channels with Large Aspect Ratios

The differential equations (2.12) can only be solved for flow

through channels with very large aspect ratios.

Velocities are

tangential to the barrel and screw root surfaces for that case,
except in small regions ne ar the flights.

The component u then

vanishes for a sufficiently large aspect ratio.

The v-component

vanishes at the screw root and is equal to Vo =7rDN at the

barrel surface
ov,fc} z.

The deri vati ve

'0 v/

~ r is much larger than

Similarly, 0 w/O z is insignificant compared with~ w./~r.

The velocity derivatives with respect to

small, as shown by (2.4).
functions of r only.

also become very

Both v and w can then be regarded as

A number of partial derivatives can now

be replaced by regular derivatives.

2.11
The differential equations reduce for u = 0, v = v (r)
and w :.:

( r)

t0

(2.1~

'b)

(2. I~c

The right hand side of (2.l4c) is a function of

pressure is not a function of r, since ()P/U'f = 0

-L~ =
'Oe

J..~ =)..A, or

R..,-tc~,I'.". 'f
,-A.A...

' .

w - w(r). Th

This

op/'br = constant. Now let

equality is only possible when

M

r.s1n~.

(Z-~)=A

13

(2. IS)

(2.1" )

Ro

where A and Bare constants with the dimension of reciprocal

time.for a

gi~en conditl~~.

The following ordinary differential equations .must be solved

to determine the velocity distributions
2
d
--u--

dr"L

'r
1

d w-

0. r"L

civeh-

+ ..L
y

Vr~

dlN"
d-V"'

::

= -Ar

(~.17

(2 .1'b)

f(."

These differential equations apply rigorously when the

channel has an infinite aspect ratio.

In the derivation of

velocity distributions and screw pump equations, we initially

assume that the equations for infinite aspect ratios can be

2.12
used for channels with finite aspect ratios.

The results

are later corrected, using known shape factors for a

prismatic rectangular channel.

2.S

Velocity Distributions

For the velo city distributions, equations will now be

derived.

The differential equation for the v-component can

be written as

Repeated integration yields

;'(' (\TV' ')

and

~ [y ~ or

IJ"' -::.

~ '("

+ '("

Co]

~ Cl+- c'&. R02. IV"

(2.2,0 b )

Boundary conditions are

r: R0

~~R,)

IJ-

I:.

'2.. 1T Ro I-J

IT=O

The ratio of screw root radius to barrel radius is introduced

to simplify the equations.
eX

Then

= Ril Ro

(2.'l2.)

The constants of integrations become

Cl

= [~R()_

ot'~I?\

C'l. = [~~ + 4-lTN/A

4"''''/~ 1(1_~'l.)

<:2.,'2.>~)

J cx'Y(I_O<'\.)

(2..'2.7.

'b)

Subatitution yields
V-: !

A.,.

r~(L)_ot7.~(~)_I&~2.e....oGl+2rtJ(Y~I?~)

l.(l_ol'\,)l

Ro

R;

\r

'(,-~ .. )

(2.24)

2.13
The differential equation for the w-component can be
written as

Repeated integration results in

and

Boundary conditions are

' ( ' &.

Re _,

r-= R, .,

W- = 0

w-=

(2.27 )

Substitution yields the constants of integration

C:!I

~(,_o('L)

4 .tv..,~

C4 ~

[I +

(I_Oc!l.)

t.,.. Ro /

~cX.J

This velocity component is then

IN:

&:. [(rl!?)'

+ (I_ol) ~ (rIR.)/.t...~ - IJ

The velo city distribution is fully described by (2.24) and

(2.)0), once the value of either parameter A or B is

kno~.

An additional equation is required to determine these

parameters.
:2.6

Scre" PumP" Egution

The velo city distribution must satisfy the condition that

there cannot be a flow rate across the flights.

That condi-

tion will yield the necessary equation for the calculation

of A or B.

2.14
An expression ls first derived for the total flow rate.

The

flow rate can be found by lntegration of w over the channel

Then with (~.30>:
. i.
-

cross section.

'L TI

1,"""or iJ. or =

; B

R; F(0< \
(2.12 )

where

The constant B is then

~=

The velo city distributions

requirement that
of r.

IX)

f2~24,

2.)0) can never satlsfy the

liquid crosses the flight for all values

Equation 2.)0. shows, for example, tht w vanishes at

r = Ro and at r = Rl' but does not vanish for most other

values of r.

This difficulty is solved in the following wy.

We specify that there cannot be any net flow across the

helical flight, instead of rigorously satisfying the condition that no liquid crosses the flight at any value of r.
The net flow across any helical surface w1th the same pitch t
vanishes since v and ware functions of r only.
Figure 2.4 shows a volume element in the channel formed
between screw root, barrel, a heli cal surface w1th lead t,
and planes

= constant and z = constant.

The flow rate

through the section AA'B'B must be equal to the flow rate

through the rectangle BB'C'C, since there is no flow through
the other surfaces.
again that

The helical boundary AA'C'C requires

(2..34)

2.15

" "-

"-

"

~~~~"?'\

\
\
\

.......

........ ,
",

I
I

,...,':.>l.. ....... ,

FIG.2-4
FLUID

ELEME~T

BOUNDED BY HELICAL SU~FACE, ROOT AND BARREL

DIAMETER AND PLANES 9 AND Z CONSTANT

. .

2.16

The flow rate through the plane z

dQ .. ct9-

1 \.N~,*'(

RI
The flow rate through the plane

rfa

dCQ: cJ ~ J, lt ~'(" ::.

Rl

constant is now

= constant is
rT<Q .

(Ro ~lfo) 01& 1

Both flow rates are the same.

v-tJ,.'f'

Then

(2.1<6 )
Af ter integration then

where

,+

k (0{' ::

2a1.'1. ~oJ

,-Q(~

The constant A becomes from (2.39)

A-

<61l'N K (o{) _

4Q.

lI5?o"!l G(0<) tO-N\.If41

G(ol)

The discharge pressure is from (2.15, . 2.16) :

Substitution of (2.33, 2.41) yields:

Pa =jJ. L

[2.
I
}]
(R.t.>-'f. )G(o() + 11 R: F("') - ~(",)~'('. _
~1Tt.J

k(o<)

4~

2.17
Here this equation is rearranged in the form of a screw pump
equation

where:

F (0<)

'2 G (ol) t~'1.lfo

T((pQ)~)~ 'l T1 -F(o() G{\ i:~lfo

4- F(at) _ 2G(o<!\ t~IfCl
The right-hand side of (2.43) contins a term proportional
to the speed N,- and another term proportional to PO/p- L.
These terms replace the drag flow and pressure back flow
rates of the existing theory.

2.7

Curvature Factors

Curvature factors FDC and FpC are introduced to compare this

screw pump equation to th at of the simplified flat plate
theory.

Equation (2.43) then becomes

(2.44 )

The choice of the helix angle

(2. 43) is very specific.

te 0

and the diameter D in

The diameter is the, barrel diamet'er,

the helix angle is the helix angle lfo at that diameter.

These

curvature factors express the influence of curvature on drag

and pressure back flow rates in comparison to similar
quantities ' of the simplified flat plate theory.
are functions of h/D

= (1-~)/2

and

PO.

Then

The new factors

FIG - 2.5 _CURVATURE FACTOR F

FOR DRAGFlOW

oc

.........tv

1.1

C'J

... ------

1
09

__ I>'<
""""

F""'"-.......

..........
~

F"""""" iIiiiiCO:

::::o"",.....!

"""""'liiiiiilIII:

>a:c::

p.. <

....", , , _

:::J

0.8

0.7

0.6

0.5

L,---!--~~-----+--+--+---t-~~~m~'>~<'=_1--1

......... 1

""""'a::c:::

"""'d

......'C:

De

0.4

"'=

0.05

0.10
--~~

0.15
0.20
0.25
h/D: Relatlve Channel Depth

0.30

''<

0.35

0.40

0.45

0.50

1.6

L-----l----~------~----+_----+_----i_----~~/~~t~~~~--~~~<'\~

1.5 ,

--

1.4

A ......L:;7'1

1.3

1.2

L,

_ .....

J---"""""""= I

---l----~--~7~S~~~~~J~7~<F~~i_--_r_r_t----r_~
~
r
I ~~
1
v

PC

1.11

1.0

k:

0.9 ,

E
, .

I::::::...................... I

::::::o:.......-L::

l"----.....::::

0.8 I

0_7

.....

0.05

0.10

0.1 5
0.20
0.25
--.~ h /0 : Relative Channel Depth

0.30

0.35

FIG. 2-6- CURVATURE FACTORS FOR PRESSURE FLOW

I:::::or:......, I

"'1

:::::::::......

-'I

0.40

0.45

0.50

r ....)

I--'

...0

(2.4S)

Figure 2.5 shows FDC as function of h/D for several values

of

~.

The influence of curvature is rather small for the

o
conunonly used square pitch ~.!D:l,
-17.65 0 ) . Th. ini"luence is lari

forlarge helix angles.

ro

The drag flow coefficients determined

by Squires (17) are identical to the coefficients F

for
DC

<fo

O.

Figure 2.6 shows FpC as function of h/D for a series of helix

angles.

The influence of curvature on FpC is large i"or the

same square pitch.

The large influence of curvature on pressure back flow is due

to the short er average channel lengths.

The channel length

in the conventional theories is based on the barrel diameter

(1.5).

It would be more correct to use the length and width

(Fig. 2.7) developed at the average diameter (D-h).

tion of these average dimensions in (1.33) yields.

Qp = Fp~ [ *(S~'llp~) ev~(~ )(.:t~JJ

This correction factor is
I - {,/D

Substitu-

..

,I

2.21

----

\,.~!
.

:. ~

T~

~cl.
?

11
~

--

IJ.J

Cl

; ~y

---

I-~\V,

t=~

:>

u:~

Ol

Ct

u.

w
:;>

1- ' 0

<t

1.,&

----

2:

~1.
I

UJ

"2

Q!

'I,
I

n..

:r

....

'2
()

~.J

:>

0
..J
~

\!)

'2

----

tl
2

'<[

!t
I

tl

111_

n..
0

...J-

\ij

'F-

t-

a
3

\!1

C\

11.1

r-

UI

I
(.)

t-

c:l

IJ}
,ft

:>
ol
J

t-

::>

llJ

.
\!)
Cl

LL

tiJ

[
..('

Of

FI G. 2.8 - CORRECTION FACTOR FOR PRESSURE BACKFLOW BASED ON CHANNEL

LENGTH AND WIDTH AT AVERAGE DIAMETER
1.6~----~------~------~------~------~----~------~------~------~----~

N
N

1.5~-----+------;-------~-----+------~------r-----~~----~----~~----~

1.4~-----+------~----~~-----+------~------~~---+--~--;-~~--r-----~

1.3~-----+------~------;-------+-----~r-

1.2~------~~-------+----------+----'~-+~~~~---------~---------~---------+----------+-------~

FP3

1.1

~------~~------~~~~-+----------t

__-=~~---------~---------~---------+----------+-------~

1.0 ~~-i======t======t======t=====f=====-1-~~--+---~---+-----I
0.9~-----+~----;---~--~~=--+------~~~~r-----~------~------+---~~

0.8r-----~----~------4_----_4------4_----~~----+_----_+~~--+_----~

0.7~------~---------~---------~----------L----------L---------~---------~---------~~------~------~

0.05

0.10

0.15

0.20

0.25
0.30
- .......~ h/D

0.35

0.40

0.45

0.50

2.2)
The correction factor Fp3 is shown in figure 2.8.

Comparison

with.' Fig . 2.6 shows that the differenee between FpC and Fp3 is
smalle

-2.8
_.-The

Influence of Side Walls

factors FDC and FpC apply to channels with

large aspect ratio. These factors must be corrected for the
correc~ion

influence of the side walls for channels with large h/D ratios.
An analytical expression for flow through heli cal channels
with finite aspect ratios was not determined because
(2.12) could not be solved.

equatio~a

The screw pump equation (2.44)

then does not account for the reduced veloeities near the
flights.

Dne can th en reason that the flow rate reduction

from this side effect is similar to the flow rate reduction in

a rectangular channel.

The side effect is there introduced

by multiplying the flow rates of the simplified theory by

shape factors FD and Fp (1.36, 1.37).
An approximation of the flow rates through heli cal channels
can be obtained when the drag flow and pressure back flow of
the simplified heli cal theory are multiplied by the correction
factors FD and Fp of the rectangular theory.
Combined correction factors are then equal to the products
of the curvature factors FDC (F pC ) (2.45, 2.46) and the shape
factors FD (Fp).
It is reasonable to base the calculation of FD and Fp for this
case on the channel width bl at the average diameter Dl = D-h.
For an infinitely thin flight land
bI:: 11 ( D - 'n ) S\M. 'f>.

(2.. ~9 )

F IG. 2 S - COMBINED CURVATURE FACTOR FOR DRAGFlOW

~~~~~---r--r---r-T-T-T-T-l

1.0NC:::

0.9 I

0.8

t
02

,< ....'C I '......'....

,,'<

0.7 I

0.6

........""C:!:

:p.....

"""

'.::

>cc:

'rl

"'cc:

'<

'"

'<"
'c

"
"

'"

'.l

"-I

L------J-------+------~------+_------~----_f------_fl'~<~--_I~~<~----~>~~~~>~~

0.5 I

0.4 '
0.05

.
0.10

0.15
--l.~

0.20

0.25

0.30

h/O - R.lative Chann.1 O.pth

0.35

0.40

....... ,

0.45

0.50

N
,t:::..

10

Fp2

0.9

0.8

20

6---,-- - 1 ---1-----t----t17~~~~~.;:o::o;;;::-+I---+------I
,"'

I ~<c::

.........<c::

'cc

0.7 I

I .x~

'n

""""'"

"3t",

HELlX-ANGLE
- 6I
i

'cc:

____
11'-

~I

c::
....

0.5

0.10

o.~

0.20

0.15
-

0.25

0.30

0.35

0.40

.....~~ h/D

F I G 2. 10 - COMB INEO CURVATURE FACTOR FOR PRESSURE BACKFLOW

>

0.45

0.50
N

N
V1

2.26
where ' ~l is the helix angle at the diameter Dl'
t-ion ot (1.2,

1~4)

Substitu-

"i.lde

(h/t '

V(I - hit:

rr (, -

'2.

+~<eo

(?.so)

hIt: tAN\.~~

The combined factors tor helical channels with finite aspect

ratios then become

The combined shape factors F

D2

2.9

and F. , are shown in Figs.

P2

Pressure Distribution in Helical Screw Channels with

Large Aspect Ratios

The calculated pressure in the screw pump channel is, for

large aspect ratios, a linear function of
partial pressure derivatives

to~

and z, since the

and z are both constant.

These pressures are at inlet and discharge then linear functions

of~.

A screw pump

wi~h

inlet and discharge connections

that do not restriet the flow must have uniform inlet and
I

d~scharge

pressures.

The calculated pressures

a~

inlet and

d'ischa'r ge do not satisfy this requirement.

This difference

results from the assumption that velocity distributions are

congruent., The case studied is really that of flow through
a section of an infinitely long screw pump.

Calculated

pressure distributions approach actual distributions in regions

that are reasonably far removed trom inlet and discharge.

2.27
The pressure difference across the flights is. of interest in
leakage flow calculations.
pressure difference at z

That pressure drop

~PF

is the

constant between points at .9

and 9 = 21\ for single . flighted screWSe

This pressure drop is

also the maximum pressure change along the ends of the screw
pump.

It is here compared with the pressure

difference~PT

generated in a length equal to one pitch of the screw (1'

t).

The pressure differenceL1PT depends on the mode of operation of

the screw pump .
The total pressure generated in the screw pump is a maximum wben
the discharge is closed.

This theoretical shut off pressure, PS,

is given by the extruder equation for Q = O.

It is a theoreti-

cal pressure that would be reached when the viscosity could be

kept constant.

From (2 . 42 a)

The back flow coefficient

(?>

(1

is again defined as

= Or I Qt>

The drag flow QD (2.43) is equal to Qp when Q = 0 and Po = Ps.

The back flow coefficient is from (2043) then also
G;)po
__
Po

= Qpo (~;: P'S.)

Ps
The discharge pressure is from this equation proportional to ~
Then

2.28
The pressure difference
;

FT"

PT' generated in one turn, is then

K (ot)
r ,r,\- N.. G(c.<)

," 11''2.

The pressur- ~difference i PF across the f1ights is from (2015)

The ' cpnstant A is known from (2.41).

That expression contains

the flow , r'ate Q, which with (2.43) can be written as

Substitution of (2.58) in (2.41) resu1ts in

The pressure difference PF' expressed as a fraction o{P1 of

PT' is from (2.46, 2'.57, 2.59)

This ratio is shown in Fig. 2.11 as function of the back flow

coefficient ~ for a screw with a square" pitch (~o ~~17.65).
The ratio is a1ways 1arger than 1 in the region 0

<f

,.

The ratio increases with a decrease of the back flow

1.

. '
coefficient .

Fig. 2.11 again shows that the ca1cu1ated pressures a10ng the
ends of the screw are far from uniform.

6 ,.

,6

6.
S

_I-

II

_I " I

II

OG pl

PRESSURE DIFF. ACROSS CHANNEL

}::> F

PT

PRESSURE GENERATED IN ONE TURN

~ ~An
30

----

PB - PC

h
::-

1TD

r - . I- - - - '

11

----i.~

==.-

P - P

..1

Backflow Coefficient f3

.2

.3

.4

.5

FIG. 2-11- PRESSURE RATIO <X P1 AS FUNCTION OF BACKFLOW COEFFICIENT, RELATIVE .

CHANNEL DEPTH AND HELIX ANGLE

...0

- Tha magnitude of A PF is strongly influenced by the mode of

operation of the screw pump.

This can be shown when the

magni tude of 6. PF for closed di scharge is compared _w1 t_~ t~a~.~

opendischarge.

The resulting ratioO<p2 is

This ratio is shown in Figure 2."12 as function of the

.....

relative depth h/D for several helix angles.

The change

in

c{P2 with a change in the mode of operation is particularly

large for screw pumps with small relative channel depth.
2.10

Energy Dissipation for Large Aspect Ratios

rha anergy dissipated in the screw pump channel is also

inf1uenced by the curvature of the channel.

The energy

equations of the ' simplified theory (1.54, 1.56) were shown

in chapter 1.

Similar equations will be derived for heli cal

channels with infinite aspect ratios.

The resu1ts are thep

compared to those of the simplified theory.

The energy is determined by integration of the dissipation
function over the liquid volume in the 'screw channel.

The

.energy dissipated between flight tips and barrel will not be

,considered here.

The contributions of that energy are identical

,to that of the simplified theory (1.6), since

t~e

clearance

ia small compared to the channe1 depth.

Th. equation for the dissipation function (19) reduces for

,,u ... v (r), w = w Cr) to:

cP -f' [( ~-7)2. +(~)~J

(2.

"'2. )

32

28

24

ex P2.

0(=

~ ..........

~ ......

f
12

2 tan2tc

~100

~,

----- ------

- - -- ---I----- --- -- -r----- r----- r--~-

........

~--

---.;..

......
I

0.1

I-----r-----

r-----~~

~-

- -'2t0 _J7.65
(t/O ~ 1)
.... ---

---

1250

h/D: Relative Channel Depth

F/G

.1-

P2. .6PF (/3=0)

20

16

6PF(~=1}

-,...

t--- ___

HELIX ANGLE _ 300

"I

0.2

0.3

I
0.4

FIG.212-RATIO OF PRESSURE DIFFERENCE IN CHANNEL SETWEEN CLOSED

AND OPEN DISCHARGE

O.S

t--I

2.32 '

The energy, EC1)dissipated in the channel per unit time, is

then

Substitution

o~

the veloc1ty d1str1but1ons (2.24, 2.30) and

integrat10n yields

Substitution of A and B from (2.33, 2.59) yields

Ee, = l~lfAA.1(~lN'l. [k'L(~)

G(o<.\

I -

r ~2._'LG(o<H~lfo/F(oI\} + J (2.64)

1 \_2..G(o(\~'eo/F(~)

~his equation reduces for small values of

0(,'1..

1_0('1-

bID to the energy

equation for tbe ,s implified theory (1.54)

The pumping energy Epe is at the discharge

Substitut10n of Q from (2.58) and Ps from (2.53) yields

"

The total nergy is ,then, '

E _E
Tc-

CI+

E Pc.-

" ..

L
"i?
'R'2.L~"l
r
k
(-<) { ~ - 2G(o<) ~lfo I F{:>t\(
I

[G(o(\

1-

0/.....

'2G{~)-t~If.Ir-(o(\(+I-~'I.

2.33
This equation converges for sma11 va1ues of h/D to the corresponding equating of the simp1ified theory, viz. (1.54) + (1.56)

The dissipated and tota1 energies predicted by this theory can

be compared with simi1ar quantities from the simp1ified

theo~

by the introduction of energy coefficients HV and HVT '

Neg1ecting the energy dissipated in the f1ights these coefficients are

~ V)...l

= ECI

(~.2.~4)

EI

(~ . I 5"4)

EI~

(~. 2."1)

lJi =

ET

(~. '2.."~ )

l2.6'f)
(2.7a

Figs. 2.13 and 2.14 show respective1y HV and HVT as function of

the back flow coefficient for a screw with a square pitch
(t/D = 1).

The cy1indrica1 theory predicts lower energies for

large relative channe1 heights.

on velocity

distr~butions

tion of these equations.

The influence of the flights

was not accounted for in the derivaFigs. 2.13 and 2.14 then only show

the trend but they are not correct for channels with large h/D
ratios.
2.11

Corrections for Finite Land Width and Multiple Flights

The previous equations were based on ' infinitely thin flights

and single flighted screws.

Corrections can be made for

2.34

..

0, ,

,
,

-"

.,'
"

r(9
~
~

~
0

'.

IU

a
IliJ

4:

0..

VI

111

-A

:
"

.J)
,

on

..
\Il

. -0.
0
0

........
~

"'0.

Q
elI
0

ti

rJ
"-

.- ,

:r

G--

.,

.
0

u.

IL'

-.:jo

0/
0

:2

J.L
\!J

"

'It
, 0

' ,'

u..
lil

~
,-

ti

"

,
o

..

'..

('4

\9

tI

lil
'-D

r-:

"

~
~

u-

2.35

00

.'

UJ.

ot

lIJ

w
l

r, .

o
lil
0

11

Q
""-

-0-

C'I

..

.0

t)

ti

0
..........

::r
UI
z
~

UJl.

3
I

.
.

C'I

cJ

<5

JI

... <r

l:>

If

ct rI-

Q..

. .

0 lfI
I- -.9

lj..

.. ..

',

C)

-\9

lJ.

2/

36

finite land width and multiple parallel flights.

Figure 2,15

shows the development of the outside diameter of a screw with

multiple flights and

fin~te

land widths.

Flow rates (2.31, 2.38)

were based on integratlons over a total angle A -e- = 2 j(

Figure 2.1S shows that

th~s

angle reduces to

for a screw with n multiple flights, each with .an axial land
!

width e.

Drag and pressure back flow rates (2.44) must, for

that case, be multiplied by a factor(l - ne/t).

The energy was determined by integration through the volume
between screw root and barrel.
by the

f~ctor

(1 - ne/tl.

The actual volume is smaller

For that case, energyequations

must be multipliedby the same factor.

Calculation of shape factors FDl and Fpl for multiple channels
and finite land widths require a correction for the average
aspect ratio of (2.50).

2.37

~.
~16. 2 \; -

NGL~

(I-'ne./t)

C'ORR~C.TIO~ (1- 'Y)e/~) Fol<. FINITE

.

LA~\) V'J1l~il-\\e\ At-JD

IY\... I'\ULTiPLE

FLlc:,Hi5

3.1
CHAPI'ER 3

3.

Leakage Flow in Screw Pumps

3.1 Introduction
Screw pumps always have a small radial clearance between flight
tips and barrel.

This clearance is necessary for assembly and

running of the screw pump.

Further, it must be large enough to

allow for thermal expansion and distortion of barrel and screw.

The magnitude of the clearance influences the leakage flow rate,
the heat transfer coefficient, and the

en~rgy " generated

in the

flight clearances (1.63)0

The flight tips drag liquid through the flight clearances in
addition to the flow caused by the pressure difference bet ween
"the sides of the flight tips.

This internal leakage flow rate

reduces the flow rate that would otherwise be obtained.

The existence of a pressure gradient (1.43) in a direction
perpendicular to the flights was not realized in earlier
theories (1, 2, 3).
neglected also .

Drag flow across the flight tips was there

Mohr and Mallouk (7) introduced leakage caused

by drag of the flight tips and at the same time accounted for
the influence of the transverse pressure gradient.
The present analysis is an extension of the analysis by Mohr
and Mallouk. " It is restricted to isothermal flow of Newtonian
liquids through screw pump channels with large aspect

ratio~.

The developed channel is again a space between two parallel

plates, similar to that used in the simplified theory.

The

, . 3.2

spa , in the clearances betw.een fli'g hts and barrel 'a lso becomes
,

.,!

a space
b:e,. iW=~.
~ parall~i
plat~s
. Again velocitiesare invest i .
....
. .
.
!

' ,l

g a~d

relat i ve to : the scre'W. " Only ,s crews with a singl flight

"

'

0'"

and un.iform channel depth, are con'sidered . .

: . f

"

To assume equal viscosities in the channel and in the .flight

. . ..
cle~rapces

is incorrect in this case.

The

~har

rates in the

clekrances are of large~ mag~itude than those ~n thechannel.

l\~st polynie.r~ have o.~lf~rent apparen~ ' viscositi.,e s at these two

,'shear ,rate , levels.

It is 'then bet.t er to use a uniform ,viscosity .

~F ,in the. clearances and a different uniform viscosit~"M--C in

;

the channel.
, In

rectangu~ar

channels, wi-th large aspect ratios, velocities are

paralle~l ~o the) barrel '~' ..e.;xcept , ~n sml.ll regions near the fl-ight

faces.

The ' velocities in the ,clearances are similarly parallel

to the brrel when t ,he land wid'th of the flight is ' large com':'
pared with. the flight clearance.

In this analysis we assume

that bo't 'h the channel aspect ratio 'and the fligh,t width to
clearance rat i o are l a rge.

v-ve

"

can. then assume veloc i t i es

, parallel to the barrel in the whole channel {lnd in all flight

clearance spaces:
3.2

Definition of Leakage Flow Rate.

'rhe leakage flow rate is the loss in , '~,low rate Jr6~ :' fl~ow
through the -clearances.
..'

In the fol;Lowing we def~ne theor~ticai

flow rates as- flow rates through a screw pump without cle'a rance.

~.

3.3
These flow rates are given by the screw pump equation (1.34)
of the simplified theory, here expressed as

Q -- Qj)T- QPT _..L

U ht
2.
Q

h~ b (]J:.)
o

0 -

\'2;k~

'0><

c:.

where the index T defines the theoretical case and the index :C
refers to the channelo
of the

pre~sure

The flow rate QT is a linear function

difference Po (Figo 3.1) for a given speed N.

The clearance c between flight tips and barrel does not change
the radial channel depth h between barrel and screw root,
according to the definition of h (Fig. 1.2}0

A pump with a

clearance c and the same channel depth h has a different

formanee characteristico

per~

The resulting flow rate Q is shown in

Figo 301 as fun ct ion of Po for the same speed No

Here t he Ieakage flow rate QL is defined as the difference QT - Q
for the same speed and the same pressure difference PO.

303

Pre ssure Distribution with Leakage Flow

The pressure dist ribution in the channel was investigated

section 1030 9 for zero leakage flow rateo

~n

The transverse flow

rate (1042) now does not vanish when liquid lea ks . across the
flightso

A different pressure distribution results when a

leakage rate is assumed o

A detail of the developed screw pump channe1 is shown in Fig. 3.2.
Here the line AIAZ is a full cross section through the screw
pump channel with a plane perpendicular to the center line of
the screw o

Points Al and AZ are physically the same point,

,

since the analysis is based on a single flighted screw.

-----Pa
F\G.~.'

DEF\N"IO/IJ O~ LE-Ak'AG RoweATE:

~-----1TD

(Q\..
~I

",
f

'.

Flow through the channel is again governed by the rduced

Navier-Stokes

equations (1.25), now in the form

(* )~
(OP ,

()'1 ) c..

~'l(A

-:)Ac.. ' ,CA

-=-

)A~

2 .....

,f'lj
J ....

,, '

,It was shown before that the pressure gradients are constants
in channels- ,w ith large aspect ratios.
to flow through the clea~ances.

Similar equations -pply

-

The pressure gra~ient ()P/Dx)F

in t4e flight clearance is identical to that in the clearance

(op/ox)C' since the change of the pressure along the line BIC~ ,
in the channel and ' in the clearance must be the same.

The

.' ,

'",-:"

pressure difference between Al and B 2 is eqUal to that between

"
A and B2' since Al and ' A2 are ,one ' and the same point. Then
2

bo(~t, +le GOSlpo(~~ )~;-(1T1)CV$ ~)(~)c:

or

"

...

The pressure gradient (QP/~x)C is identical to the pressure

gradient (1011), derived in the simplified theory

OP)

(~ r-

Equation (3.3) contains two unknown transverse pressure

gradients.

An additional equation, required to determine these

1

gradients, is obtained by equating the transverse flow r ,a te ip

the channel to the transverse flow rate in the flight clearance.

3.6
The flow rate leaving the channel across BleI is per unit flight
length

a c::.
-.l

_12. 0: h

_ .!{.

_J

I 2)A..c:..

1e )
)~

(),5)

A similar flow rate entering the flight clearance is

calculat~d~ ~ but

The gradient can now be

cumbersome.To

s~mplify

the results are rather

the analysis let

(op/o'1 )F
C>P!()X )c.

(0 ~ /'1 '1

) c ...

C;'h,>< )c.
Introduce further the 'oack flow coefficient ~ T for the screw
pump without clearance, defined as

This Cfficient is from (3.1)

~T

hL.. ,
= . ~ l-ic fAc.

(~)
,>'X

c:..

The barrel velocity component Vo can, with (3.10), be expressed

as
Itt
IJl
_
h2.+avv-.'e.. (~)

y ~ ,~ U. o

o..tY\., T 0

~ T "uc..."x

c..

(~.

11 \

SubstitutiIlOf (3.8, 3.11)in (J.7) yields, af ter rearrangement

3.7
Equation (3.3) can be expressed in a number of geometrie ratios
using the relations

bo

== (t -~)

voS

11 D = -I:; l-t~ 'f

(3 , , ~ ",,-)

Y;(J

(3.13.

b)

The result is
otc --

eltI_

e/t O<f:

( 1-

e/t) +~lfo

Let now

t == .elf-

0-

(3, ,-\

\.f..h )3 (~
)
~F

(3.lb \

Elimination of o(c from (3 . 12, 3.14) yields the expression for

O<F
o(!=

+ ( I - ~ )(

[ ( I-

Similarly
0( c --

L ) cr

1_

c/h ) +o..-v..1-Cf4/f'1T

-+ [ ] +~ y

<)

t ( ,- Cl" ) +0/1/1.7.. CPc Ir T - 0[Cl,- () cr- + E ] +0-.. ~

<)

A detailed discussion of the pressure distribution will not be

given since it does not differ a great deal from that in a
pump without leakage when the leakage flow rate is smalle

The

above results will now be used to calculate the leakage flow

rate . QLo
3.4

Leakage Flow Rate

The analysis of the leakage flow rate is simplified with the

introduction of the control region AIA2B2 of Fig. 3.2, where AIA2
is the full developed cross section through the screw pump.

The

flow rate leaving the control region through that cross section
is equal to the total flow rate Q through the screw pump.

3.8
The sum of the flow ratesentering

the region across the

boundaries AIB2 and B2A2 must be equal to Q. Now it is much

simpIer to determine Q from the separate flow rates across AIB2
and B2A2.
Each flow rate can again be regarded as the superposition of a
drag flow rate and a pressure back flow rate.

Theequations for

these drag flow and pressure back flow rates are quite similar
to those of ,the ,simplified theory.
The control region ,is further subdivided into the channelregion
,

AICIBI and the flight region CI A2B2BI

x-direction are investigated first.

Flow rates in the

The channel drag flow rate QOC is identical to the theoretical

drag flow :rate

The channel pressure back flow rate QpC is similarly.

~rag

The flight clearance

flow rate QOF and the pressure back flow

rate QPF aredef'ined as flow rates across the line BI B2 . ,:r'hey

are

Q1)~

U o c (e. co,S ~o)

(3.20 \

'.

.,
~ (e Cos(" \ -L
'-t'l'F:: \ "2..
.
ro) M F

..

('1.f.
\
"'j)')(:) c.

Next we investigate flow in the y-direction.

(3. '2,

The flow rates

that cross the line B2A2 are

:l)r

Ql'F(1

tLfo

c."!>

\'2r

('Ti D ca~ V'... )

(11 D CoSf.) f'tf

_I (YJ?.)
0'1 F
<J

(3.2,,\
l~. t 1)

).9
The total drag flow rate entering the region across both
boundaries AIBl and B2A2 is then

Qoc:. + QD~

Qc =

- G Df '(

Substitution of ().la, ).20, ).22) and

V0 -

tan

U0

11'

D t

'f 0

an 'f

b o = (t - e)

cos~o

yields the total drag flow rate

t ( h_ c.) b

Ot) =

(1. :-)

lAa

This equation shows that the drag flow through a pump with a
clearance can be obtained from the equation of the simplified
theory when the channel depth h is replaced by an effective
channel depth h - c.
Next it will be shown that the flow rates resulting from the
pressure gradients can be expressed in QTD.

The back flow :

coefficient is first used to eliminate the pressure gradient.

( ~)
'"0x c.

From ().lO)

:.

Substitution in (3.21) yields with (3.16)

(3.2'7 \
A similar substitution in (3.23) results in

The leakage flow rate QL isthen _

Ql."::: QC =

b
C

QPT -

Qo;- Qpc ~ Q: - Qrf'-d'

~o -+

QPi= -

Gt'F4-

).10

Both QPF ().27) and QpFY (3.28) now contain the theoretical

dr~g

flow rate QOT as a factor. . In leakage flow calculations it is

advantageous to express the leakage
drag flow.

~low

as a fraction of that

This fraction is

1'C..o

('3.30\

GL

~ DT

Substitution of (3.27, 3.28, ).29) in ().)O) leads to the

leakage flow coefficient

This coefficient is from (3.15, 3.16) a function of the ratios

c/h, e/t, the helix angle ~o' the viscosity ratio and the back
flow coefficient ~.
It consists of a fraction kc that does not change w1th the mode
of operation

~+
h

0-

(I - <-lh ')

cr(I-t.)-t-

and a fraction that is proportional to

t.

0T

The influence of different parameters on leakage flow can

effectively be shown in a dimensionless performance diagram
(Fig. 3.)

which relates the dimensionless flow rate

to the dimensionless pressure

3.11

\.0

o.~ I------.:~~----+_---- -- - - -

ol:> 1------+---....--

- _. -

.-- ----- - ---- - - - 1

I

"1

fT

I" ~'-r
I

o.~ f-----+-----+---~'--"'
I -----t--\

"I

0,'2.

I----t------:..----~_+_I-'~o

o.~

0.6

0.4

0.'2.

1.0

PST is the shut-off pressure for the pump without c1earahce.

The dimension1ess chnracteristic for that pump 1s the straight
line

(~.%)

The characteristic for a pump with clearance is then

t -=

T -:

Q\..-:-

\ -

~T

fc<)

Q'DT

Substitution of (3.32, 3.-33) resu1ts i

= (\ - -f<c) - ~T

( \

+ -kf.>o)

3.12
The dimensionless characteristic of a .pump with clearance is
then also a stra.ight line.

It intersects the f-axis At: a poin'c

l-kc' the (3T axis at a point ~ T

(l-kc)/(l+k~).

The coefficients . kc and k(~o are functions of c/h, e/t, jvtC/t-F

and ~o.

A number or dimenaionleaa diagrama were conatructed

o

for ~~= 17.65.

The other parameters were one at ~ time

varied in eachdiagram.
Figs. 3.4 and 3.5 show the influence of clearance on performance.
Both diagrams were constructed for constant viscosity ratios.
A viscosity ratio was previously introduced to account for the
non-Newtonian character of polymers.

Now, a change in the

clearance ratio c/h will then also change the viscosity ratio.
The influence of screw wear on performance can be predicted
from these diagrams.
Fig. }.6 shows the influence of the viscosity, while Fig. 3.7
shows the influence of the land width ratio a/De

The influence

of the width appears to be smalle

The diagrams show that leakage flow can become negative.

This

is only possible when the inlet pressure is much larger than

the discharge pressure.
in screw pumps.

Such an operating condition is uncommon

It can occur in extruders when the melting

zone generates a large pressure which is partly throttled in the

metering zone.

3.13

.b

0:4

0:2

-0.4

-o.~

0.'2...

o.~

\,0

1.0

<.po = 17.<O s o

e/ D = 0, I ) ,A...o<-:
f

3.

/~ ~

o.b

0.4

0.'2.

-o.~

I. 0

/.0

~o=I7~SO

c/h:. 0.\
Q./t;,.: 0.\

O.r::,

0.2.

--~ (1T

_0/2

Flc:,. !:..c:,.

1.0

0 .2..

INFLVt::NCE. Or VISCOSlT'1 IcATIO

ON

L~A'O\G..

1.0

<{oo=17, br;o

C/h: O.

J.::~F:3

0/.,

0.'2.

-0:2..

0.'2

O.b

o.~

10

3.15
3.5

Leakage Flow Rate and Heat Transfer

So far it was assumed that all velocities everywhere in the

channel are parallel to the barrel.

That assumption is reason-

able, except in small regions near the flight faces.

The : exact

solution for the transverse velocity distributions in a

rectangular channel results in a circulation, schematically
shown in Fig. 3.g.

The projections of the stream line are

shown here as closed curves for a screw without a leakage flow

rate.

These stream lines are parallel to the barrel in most

of the channel.
The stream

linesmu~

be different for a pump with a leakage

flow rate, as shown in Fig. 3.9.

The leakage flow rate through

the flight clearance at Al must be the same as that through the

flight clearance at A2 That leakage flow rate occupies the
space between the barrel and the connecting stream line AlA2.
This stream line is parallel to the barrel, except in the : small
regions near the flight faces.

The liquid leaking through the

flight clearances is then contained in a thin wall layer with

a thickness c*.
channel width.

The layer thickness is uniform in most of the

Later it will be shown that this thickness c*

differs from the cJearance c.

The liquid contained in the wall layer flows in the proximity
of the barrel and does not take part in the circulation.

The

stream line AlA2 separates the liquid in the channel in a :

circulating stream and a stream that always remains in the wall

3.16

FIG.~.t6_ SC\4E~Ai\C. D\AG~AM Of- RAwS\l~Se

C \RCU L t\\~N \"J CH~NNE L. W \n40 v T Cl.EA.RA.tVCE

1=\6 .3.9 -

FORfV\ATIO~ 0\= \..A'1ER. C

NEA~

BARREL

IN SCRE WPV MP W\T). 4 LfA.K .t..GE

3.17
layer.

In other words, once each revolution the barrel is

wiped clean except for the wall layer.

The velocities in the

wall layer are substantially equal to the barrel velocitY.1

This wall layer plays a dominant role in heat transfer to the
liquid in the channel.

In a heated screw pump, heat is cqn-

tinuously transferred to this wall layer.

The innerrace AlA2

of the wall layer is exposed to the circulating flow rate jonce

each revolution.
Jepson (20) calculated heat transfer coefficients for the
following model.

The film that is stagnant relative to tr,e

barrel can be treated as athermal resistance without

appreciable heat capacity.

In the . time required for one

revolution, this film is exposed to the circulating flow.

The

time interval for such a cycle is so short that the liquid in

the channel can be treated as a slab of infinite thickness.

The

amount of heat transferred to the slab in that interval is then

calculated by Jepson for a given wall temperature and a given
uniform initial slab temperature.

This amount of heat can be

converted to a heat transfer coefficient.

Jepson calculated heat transfer coefficients as a function of
the screw flight clearance and as a function of the number of
wipes per minute.
It was shown in the previous section that the leakage

flo~

rate

QL' for a given screw pump geometry and a given viscosity ratio,
changes with the mode of operation.

One must then expect that

the thickness of the wall layer also changes with the same
naramet er s.

3.18
The intent is not to give a detailed analysis of heat transfer,
but merely to point out that heat transfer coefficients cannot
be based solely on the nominal flight clearance.

They must be

based on the thickness c'* of the wall layer and, therefore, are
influenced by the geometry of the screw, the viscosity ratio and
the mode of operation of the screw pumps.

The magnitude of the

wall layer thickness is discussed in the next section.

3.6

Leakage Layer on Inside of Barrel

In heated or cooled screw pumps, - the thickness of the wall layer

is also influenced by the temperature distribution in the
liquide

Here these temperature distributions are ignored.

We

again assume different viscosities JA-F in the flight clearance

and/t v

in the channel to account for the influence of shear

rate on

viscosity.

The same viscosity f"-c is assumed for the

circulating liquid and the liquid contained in the wall layer,

while that layer is in the channel.

A different viscosity would

otherwise invalidate all previous equations.

The wall layer thickness is derived from the transverse velo city
distribution.

That distribution is obtained by integration of

(3.2b) for the boundary conditions

lY( d,a);::
l)( ';1,h)::

Then

The leakage flow rate qc per unit flight length is from

and (3.26)

(3.5~

3.8b)

------------~------~---------------------------------------------------,

3.19
This flow rate can also be expressed as
(~.41 )

The coefficient ~y is a transverse back flow coefficient similar

to the back flow coefficient for flow in the channel direction.
Equation (3 . 39) can be expressed in analogy to (1.51 a) as

\l;, [(~)(I_1>~~) +

V"'=

"~~(~)J

(3.42..)

The flow rate in the barrel layer is

o

:J. c

::

~
\Jct.

JI-

Yl_

c.lil

Substitution of (3.42) and integration yields

c* 2 c")
cl/(

<ie :;:'

~ h Lrh (

f} h

(3_ '2 ~

:je

Theresults of (3.41) and (3 . 43) must be, the same.

Cl!( (C~

~d- :: h

I-

This equation

~as

co:

:2 -

laye t: thick~ess

The remaining roo,t ' is

(3.45-)

from (3.17 b)yields the relative barrel

,

0-(

C \
--= -

c*

2h

i (~-I) = ~. [t;;~c -: I]

~c

.,. ,,

2( 3 -

a double root at c* = h.

Substitution ,of
,

cf<

h ) - ~~ ( h)

Then

($r 11:~<to) + E (cl"') "

1- ,E +

E ( I - cl h) - ~,.

"

0"' /

-\;~tfc

The re;I.ati ve t ,h ickness is expressed again in the geometrical

parameters "c/h,
"

'

e/t,
,"

flow ,coefficient ~ T.

:and <p"
0

the .'vi'scosity ratio, and the back '

,,

'

.
'
. ', '

A,riumber of cases were a.iculated for a

helix angle of 17.65 0 (t/D'

'1 ) ~ '

n~ of- the parameters was

varied in eacn case - the others kept constant .

3 '. 20
The operating conditions are varied in figs. 3.10 and 3.11 by
varying ~ T.

A Newtonian liquid is used in fig. 3.10.

The curves

show"that the thl.ckness c* is significantly influenced by the

mode of operation of the screw pump for relative clearances
larger than 0.04.
dotted line c
~all

Jepson's heat transfer model is based on the

c*.

For normal operating conditions, the actual

layer thickness is smaller than the clearance used by

Jepson.
Fig. 3.11 shows similar curves for a viscosi ty ratio

cl? F

5.

The influence 'of the mode of operation now becomes significant

at a smaller ,'relati ve clearance c/h.
The influence of the remaining parameters is shown 'in Figs. 3.12
and 3.13 for a back flow coefficient ~ = 0.2.
the viscosity ratio is shown in Fig. 3.12.

The influence of

Thicker wa,11 layers

"

result for larger viscosity ratios.

Viscosity ratios smaller

tQan 1 are included to show the trend for cooled screw pumps.
Finally, Fig. 3.13 shows the influenceof the land width for a
viscosity ratio ?c/ ~F

3.

The layer thickness does not

change a great deal with a change in e, as would be

from Fig. 3.7.

predi~ted

>

"

Fig. 3.11 indicates "that the calculated values 'for c*/h can
"

become negative.

This ,occurs

of (3.40, 3.41) becomes

whe~

negativ~ "

the transverse flow rate qc

-'which is only possible ,

when the pressure gradient in' the 1;iransverse directl.on is

so large, that ~ y

> 1.
-',

."

3.21

O. \"

r - - - - - - - - - - - - - - - -__-----+----....,r--....--,r----~
"\SCO~,T'1 RA"na Jv-C/)AF: \. 0
\-lEL\'>< A"-'GLG \.~5<'

F\. \G::~'" W'''t>T~

eAlO

Q,/o :

o. \

0.\0 ~---~----~----~----~~~-~~--_,~

o.oS'

O.O'2.~

FIG.~. 10_ 2f\...ATI"I

O.o1~

WAL.\..

I~ ~I..VENC."

elF-

O. \2.~

0-\0

\.A"1t:~

TH\CkNES<i

o~

,,",OJ)E

O.ISo

C~h Fo<a).AC/"..v.f=I.O

orE I\TIO~

O.\~ -r_--------_.----~r_~--7r------~-~

~c /)A.F

= 5'.0

~o = 17

"5

Q..lf:>:. o. \
0.\0 ~---~----_+---~~~-~--~-~--~---~

O.Osr-----+---~~~~--_7~------~--------_r------~
_ 0.2..

o.o2~

o.o~

O.o7~

0 .10

~ I2E L.AT)'-Jt;. l P C\.~eAhJC~

FIC. ~ .. I' _ eE\.A'''Je

WA\.\..

cf h

LA",t~ -n-t\c:~~s c/h

INFL.u&NCS ot: ",",ODE Of: o~~ItATIO'"

a.l'S-

O.IS"

FOa.,Me. /f1J:=

s;..o

3.22

O,I~ .----------------r----~-~-___:lr_--__z71
BAO<FLO\N COEFF. ~T:: 0.20
HELIX ,6....,GL \.po = n.~sCt
F\..IGo\-\'T' WIOT\.\ RA\o '<-/0 - 0.10

O. 10 +------.------,------+-+-~71f7Co..--7'--t_7"-_:;;rl

O,O~~----+_---~~-~~~~~__~-_;-----~------I

o.()~

0.125'

0.10

F\c:,.~.v.2.- II-)F\.VEt.JCE Of \lISCOS'''''1

RATIO

REL..A"I"~ W\,.L.. \..A'1~R. ""I(J(N~SS.

}Ac./)A,.

0.15'

t=oR., . ~trO.?-) E/O:O. ,

0.\5' -----------~----,..---__r-""lr--__r~~r_---~
~ = 0.20
\.pO:' 17. c:, SO
VISC.R#\lO

jA-tJ)J.r: '3

0.10 ~----..:.--------t-----hf-___,~*~T---___lr_---____1

O.os~------~~------~~~~~~--------_r--------1_------____1

0.015"
FI c::,.

0.0>

0 .o7-

0.10

O.I'2r

0./'>0

'3. \~- r~ FLVEt.Jce ($. 12E.\..ATI\J~ r\.\G~ w IO"'~ "IDot-.)

R~ l.~)\J~ 'NA\..1.. ~l\'1(.1'2. "NI C\(.IIJESS to~ ~T= 0.2) h(./Mf ="1.

3.23
Negative wall layers cannot exist.

Ins~ead,

a layer is formed

on the screw root surface as shown schematically in Fig. 3.14,

Some of the liquid is still dragged by the barrel through the
flight clearance, but a larger quantity flows back into the
channel.

The equation for the relative wall layer thickness (3.46)

does not apply to negative leakage flow rates.

That layer

thickness c*l must be calculated from (3.41) and

c~

<lel

v-o. t

(3.47 )
I

o
Substitution of (3.42) and integration yields

Ci CI

~ u;, h (Ch\'- [I - ~'1

(L 2.

(3.48)

'() ]

This flow rate must again be equal to the one given by (}.41).
The resulting equation has one root at c*l
are

((\_I)

h.

The other roots

v1r~-')(9~~-')
4 r~

-The transverse back flow coefficient ~ y must be larger than 1

for this case.
values for c*l.

Only the plus sign in (3.49) results in positive

The occurrence of a screw root layer will be

th exception rather than the rule, since it requires some

combination of a large clearance, an inlet pressure higher than
the discharge pressure, a very small land width or a large viscosity ratio.

The relative screw root layer tends to be large when

it occurs, since the velocities are small near the screw roqt.
A number of examples are given in Table 3.1.

3.24

( SCREw iooT LA'1~~)

FIG.

?>.\4 _

CHEMA lC.

ce.e. w

O/AC;RA"1

$\4 oW I "-Jo,

RaaT Lr..-.~t:n. c.:~

Table 3.1
Relative Screw Root Layer Thickness
for ~ 0
~T=
)J..c

/P F

-0.8

17.65, e/D = 0.1

-0.6

"-0.4

-0.2

0.1217
0.0094

c~/h

c/h

5
5
5

0.150
0.125
0.100
0.075

0.6181
0.5052
0.3625
0.1997

0.5386
0.4230
0.2849
0.1289

0.4119
0.2982
0.1669
0.0053

3
3
3

0.150
0.125
0.100

0.5046
0.3839
0.2453

0.4188
0.3000
0.1655

0.2856
0.1682

~,------~--------~~~~~~~--------------------~------------------,

.4.1

CHAPTER 4

4.

End Effects in Channels with Large Aspect Ratios

401 Introduction
Screw pump channels are in conventional theories (1, 2, 3, 4,

6, 7) replaced by prismatic channels with rectangular cross

sectionso

No mention is in these theories made of the

bounda~

conditions along the oblique ends of the screw pump channel

(Fig. 4.1). -Congruent velocity distributions are assumed in all
cross sections perpendicular to the center line of the screw.
The pressure is in the simplified flat plate theory a linear
function in x- and y (1.44).

The pressures along the oblique

ends, predicted by that theory, are then not constant, except

for one special combination of drag flow and pressure back
flow.

. The same applies to the rectangular channel an!lysis.

The screw pump must have uniform pressures along these oblique
ends when the inlet and discharge ' connections of the pump do
not restriet flow.

The conventional ahalysis does not satisfy

-~ the actual pressures along the channe(:1nlet and discharge endso

Flow through screw pumps is in this Chapter studied for uniform

pressures along the oblique inlet and discharge ends.

The

resulting drag flow and pressure back flow rates differ from
those determined in existing theories.

These new flow rates

are again expressed as fractions of the corresponding flow rates

of the simplified theory.

End effect correction factors are

defined as the ratios of the new flow rates to those of the

simplified flat plate theory.

This analysis is restricted to isothermal flow of very viscous

incompressible liquids with uniform viscosities through
channels with large aspect ratios.

Large aspect ratios are

used to keep the problem two dimensional.

The majority of

screw pumps and metering zones of extruders have aspect ratios

in excess of 10.

The results of this theory are then useful

for most cases.

Acceleration and gravity forces can be neglected for slow flow
of very viscous liquids.

Velocity distributions are again

studied relative to the screw.

Leakage flow across theflights

is neglected i n this analysis.

4.2

Geometry of Screw Pump Channel

The helical channel with large aspect ratio is, in the simplified theory, replaced by a straight prismatic channel with a
redtangular cross section.

The same prismatic channel is used

in this analysis, now however with the specific requirement that

the ends of the channel make angles ~o with the channel
direction.

The schematic development of this channel is shown

in Fig. 4.1.

The barrel becomes a flat plate that moves with

a velocity Wo

11' D N over channel and flight tips.

We must then determine the velocity distribution in the region

AA1Bl B:. for uniform pressures along the inlet and discharge
ends while the barrel moves with a velocity Wo in the direction
shown in Fig.

4.1.

4.3

~o

I
]

uJt.!1

:> \.9 Q
() <;!!

d)

"2

UJ
~

IJ}
~

:; < U)
caI

liJ

Q!

4"
:r

O~

J
I.U

Cl

VI

:>

Cl

af

\u'

0
u

j
LU

\,

\,

...J

\-

l:I"

"

:i

u.

j...-OQ

U-

-~~
~-~~

~->

4.3 Average Velocities and Velocity Potential

All velocities in a channel with a large aspect ratio are
parallel to the barrel except in small regions close to the
flights.

These regions become very small when the aspect ratio

is very large.

The velocity component w perpendicular to the

barrel vanishes everywhere in the channel except in these small

regions.

In this analysis we assume such a large aspect ratio

that these regions become insignificant and w vanishes in the

whole channel.
Steady flow is, in the absence of acceleration and body forces,
governed by the equations of motion
~
x

~
';}

or
'01:

V\A..

(4. \)

;U- V"IJ'

(4.2.)

(4.~)

The component u vanishes at the screw root, z = 0, and is equal

to the barrel velocity component Uo at z = h.
u~z

The derivative

is then much larger than the derivatives with respect to

x and y

In a similar way th deri vati ve d v / zis much larger

than y/dx and 6v/by.

Equations (4.1, 4.2) are then approxi-

mately equivalent to

(44'
(4.5")

The pressure is independent of z while the terms on the right

.. ,

'

hand side of both equations are approximately independent of

x and y.

Th equations can then be integrated.

For eX8lllple
lA. ::

(-L
h)
2/Jv ())(.

j!z

+ C, ? + C...

(4.')
,

Boundary conditions are

z

0,

- ' h,

Uo

The constants Cl and C2 of (1... 6) can be. e.aluat'. cl yi.1d1nc aft

equation for u .in the form
IA

= (~

~)[

? ( i! - h

)1 +

Lt.

("2/ 'n)
,,

Similarly

(4. Q)
The average velocity components and 'y are now d.t.~1ned by
integration of (4.7, 4.6).
lA.::

.lS~a~
h
0

Similarly

Lr: -L.
h

For example

:_

12}A-

r 1T0-2 : _....u:..
Jo
\2.jA

~
x.

~2

(4.q\

~ +~

o~

2-

These mean veloc1t1es can, following a principle u d bJ

Hele Shaw (6),

b~

regarded as the veloeiti tor a .eloc1ty

potential def1ned by

lfJ

h'2. p _ 1.. ~lo)(

\'2fA--

'2..

J..
'2.

v: ~
0

Cl

The mean veloeit1es are then

Th. equat10n of cont1nuity for the mean .eloo1t1.. CaB ba

written as

Substitution of the velocities from (4.12) yields

nh<.~ +

o"2.~

Q~'l =

The velocity potential satisfies Laple's equation.

Calcula-

tion of the mean velocities requires the solution of LapLace ' s

equation for the proper boundary conditions.
Boundary conditions are formulated for the frame of reference
of Fig. 4.2.
y

The mean velocity

v must

vanish along the flights

bO because no liquid can cross the flights in the

absence of a leakage flow rate.

Then

In Chapter 1 it was shown that velocity distributions in

I

rectangular channels with large aspect ratios differ from those

of the simplified theory only in small regions near the
flights.

The

the flights y

s~plified

theory prediets a uniform

U,

even along

bO.

The average velocity vanishes along the flights in the

rectangular channel theory, since a viscous liquid cannot slip
over the flight faces.

In that theory it becomes equal to the

average velocity of the simplified theory at a small distance

from the flight faces.

This distance, which is of the order

of h, will in the analysis of end effects be ignored .

flights we then accept

At the

r O.

The obliqua ends are described by the lines

>< = +

I tC\.M., tea

(4. IG )

4.7

JIN
/I

l-

\I

(!

a
-

VI

(3
11

if)

o
lA
2

F,i

-t:

~
~

"
.:f';1

3<;)1---

-3

~
~

I
I

11

SO

3-{":r

rot'

d
I
11

><

Let PI and P2 be the uniform pressures at inlet and discharge.

The potentialQ;2 at the discharge end is then

Y;2.. = h'Lp2.

\ lA

12fN-2

CA.
0

(Uo v: )

-'2~t~"e,,+o

Along the inlet similarly

'fI, :::
The velocity potential has singular points at the four corners
A, Al' BI' and B.
The velocity potential could be determined for the general
boundary conditions.

It is more useful to determine the

influence of the ends on drag flow and pressure back flow

separately.

We will show that the general case can be regarded

as the superposition of several simpIer cases.

4.4 Superposition of SimpIer. Cases

The general case can be regarded as the superposition of
simpIer cases.

These cases will be so selected that the drag

flow and pressure flow end correction factors can be determined

immediately from the solution of two of these cases.

Use is

made of the superposition principle previously discussed in

section 1. 3.6.
The general case is here separated into three cases, each
satisfying the boundary condition

?:J'f/la d- = 0 along the flights.

The boundary conditions for the three cases are shown in the
following tabIe.

TABLE 4.1
Case No .

Uniform Pressures at
lnlet
Discharge

Barrel Velocity
Components
Uo

Vo

Uo

Vo

Uo

Va

SUM

! (Pl+P2)

i (Pl+P2)

-PO=(PI-P2)/2

(P2- Pl)/2=PO
P2

Pl

The first case was so selected that it is identical to the drag

flow case of the simplified theory.

Pressure gradients then

vanish yielding a uniform pressure in the channel.

The

pressures along both oblique ends is then equal to that uniform pressure

This uniform pressure was here equated to the ;

average pressure.

The velocity potentials (4017, 4.18) along

the ends become

The velocity potential in the region is further from (4.11)

(IJ

'f' -

__
I

'2.

,U

X
0

The average velocity component is =

while

v variishes .

Uo

The drag flow rate is

ODO

==

lo

bo h

The velocity potentials along inlet and discharge for the

second case are
Then from (4.11)

4.10

Fig. 403 shows schematically how the liquid flows through the
channel when the barrel moves with a velocity V00

The barrel

drags liquid into the discharge end and removes liquid at the
inlet endo

The resulting flow rate QDl tends to reduce the

total flow rate.

This flow rate must be proportional to Vo

while it is independent of the pressures PI and P2.

The flow

rate for the last case of TabIe' 4.1 is clearly independent of

Uo and V00

QDO and QDl are then the only flow rates that

have the character of a drag flow.

The _total drag flow for

channels with oblique ends is equal tri

QDO + QD'"
Calculation of this drag flow requires the solution of

QbE

Laplace's equation for case 2 only, since QDO is known (4.22).

CO"-J\JEtJTIOt-J

Qt)\ >0

Fol<. GDI :

\N H~N IN DII<E-c.IloN of

'.

F\G.

Dut;

4.3Tc

SC\..\I'YlATIC

FLow D\STRI'BUTIOt-J

~~RREL VEL.Oc..\T'1 CO"",,PoI.JEl-Ji"

-/

u;, AL.OI\JE

0 DO

4.11
The last case of Table 4.1 corresponds to flow through a
prismatic channel with oblique ends and stationary walls from
uniform pressures along inlet and discharge ends.

The

velocity potentials along the ends are from (4.11)

"+' I ; :

h'l.

~2. = -' (p - P7.)2.

'2.

I?.~

= -

h t?o

(4:2.6)

\2.}A-'

The inlet and discharge pressures were

50

pressure vanishes at x = 0 and y = O.

The resulting pressure

~ack

selected that the

flow rate QPE differs from the back flow QpO of the

simplified theory.
pressure gradient

The latter was based on a constant

p/x,

while op/~y vanished.

Case 3 and the

corresponding simplified pressure back flow case must have

identical inlet and discharge pressures at y

O.

Comparison

of QPE ahd QpO yields the end correction factor for pressure
back flow.
The analysis of the general case has now been reduced to the
analysis of the simpler cases 2 and 3.

It will later be shown

that the results of case 2 can be predicted from the results

of case 30
4.5

Symmetry Considerations

Symmetry allows a further simplification of the problem.

The

drag f19w velocity distribution induced by Vo is treated

firsto

Consider two points Aand B (Fig. 4.4a), symmetrically

located relative to the origin on : a line through the origine

The mean velocities uA and u B in A and B change signs when
the barrel velocity component V is reversed (Fig . 4.4b ).

4.12

A
ti: "
--~

......

~-+---~--

.B

,"

(0.)

~-o
>'1 A

M:o

Va

0'1

(b)

R-oTA"O/..J

AI<avN.D
r-AX\~

~::a
()~

B
X

Q.-o
~'1

CC)

'.

4.13
Subsequent rotation of the whole system around the z:-axis
through the origin through an angle of 180 0 interchanges the
points A and Bo

The boundary conditions after this rotation

(Fig. 4 . 4c) are identical to the original boundary conditions

(Figo 404a}o
be the same

The magnitudes and signs of uA and u

must then
B
The mean veloeities vA and vB 'must, for the same

reason, be identicalo
The gradients of the velocity potential must then from (4.12)
be equal in points that are symmetrically located on a line
through the origino
I

The velocity potential differnce between Band 0 is equal to

the similar difference between 0 and A.
points

~and

This also applies to

D located at inlet and discharge.

Then

(4.27)
From (4 23), however,
0

since
Substitution in {4027} shows that the velocity potential
vanishes at the

origin ~

Thus, any two points diametrically

located on a line through the origin have equal but opposite

velocity potentialso
Lapl ace's equation then requires solution in half the channel
region onlyo

The additional boundary condition

4.14
can be used to determine that solution.

similar symmetry

.A

applies to the pressure back flow case since the velocity

potentials at inlet and discharge (4.26) have equal and
opposite values.
4.6

End Correction Factors

The drag flow rate QnE for channels with oblique ends can be
expressed in the corresponding simplified drag flow rate Qno
through the introduction of the drag flow end correction
factor FnE

Then

ODO FJ:)E

ODE -

From (4.22, 4.25)

Fl)E

where

OD\=

l+

201>\

Uoboh

bo

bo

hS2~ 01'1

==

ba

hf

_bo

-'2-

(~)
'0'1<

d~

)('=0

2.

Introduce now the dimensionless velocity potential

and the dimensionless coordinates

The reduced potentials along the inlet and discharge ends are
from (4.23)
"7
z:..DI

while

'")

Z __
1)2. -

1Yo~
2. \Jo bc)

__ .lU
-

Zn!oY vanishes along the flights.

2.

cr

4.15
Further
Substitution of (4.31, 4.32, 4.33) in (4.30) yields
J...

Fp =

I - 2.

1:Qm.~o

L(!~~ )X:: ~
-+-1,..

'2.

The regions, expressed in the dimensionless coordinates X and

Y, are identical for geometrically similar screw pump channels,
while they have the same dimensionless potential Zn alongthe .
boundarieso

The integral in (4035) is then a constant for

geometrically similar channels.

The end correction fact.o r is,

therefore, a function only of the factors that describe the

dimensionless region.

It is then a function of the helix

angle ~o and the ratio a/boe

The pressure back flow rate QPE can, in a similar way, be
related to the simplified pressure back flow rate QpO.

The

pressure back flow end correction factor FpE is given by

The simplified pressure back flow rate must be based on

pressures

PO

at points y = 0, x = + a, where Po =(P2- Pl).

For that case

The simplified pressure back flow ra'be is then

4 . 16
The flow rate through the channel with oblique ends is
+~

QpE '"

Jb~

~ t~'1

(4. 3 ,,)

-7:

Let now

~f>

--

Lp /( tth
I~~ \)

+i

then

FpE

i: L.(~~Lol ~
"-

At the oblique ends from (4.26, 4.40)

rPI

=-

~ . 4'2.)

LZp2. = - \

while Zp/'Y vanishes along the flights.

The integral (4.41)

is agin a constant for geometrically similar channels.

Thus,

the end correction factor FpE is also a function of the helix

angle~ 0 and the ratio a/b O

4.7 Numerical Method

The

~nalytical

solution to the problem was not obtainedo

The

method of finite differences was used instead to obtain

approximate solutions.
The factors FDE and FpE must -be determined for a range of
helix angles' and a large number of a/b O ratios. It is not
essential to know the full solution of ~J through all regions
when only FDE and -FpE are required.

4.17
A numeri cal method was devised for the solut ion of large

numbers of partial differential equations for problems with

the following propertieso
a

the region, for which the desired property of the differ-

ential equation must be calculated, is one of a family of

regions.

The family of regions consists of regions .with one

identical dimension, while the other dimension increases

systematicallyo
b - the required property can be expressed as a geometrie
property of that differential equation in that region.
The end effect factors (4.35, 4041) are typical examples of
geometrie properties of a differential equation in a region.
The family of regions in that case consists of channels with
the same helix angle ~o and different ratios a/b o.
The shape factors Fn and Fp (1.36, 1.37) for flow of viscous
liquids through channels with rectangular cross sections are
also geometrie properties of Laplace's and Poisson's equations
respectively, now in a family of rectangular regions with
different aspect ratioso
The numeri cal method is explained and illustrated in Chapter 6.
There it is shown that calculation of the required gmetric
property does not require calculation of the function through
the whole region.

4.18
4.8

Functional Relation Between FDE and FPE

A simple relation exists bet ween the end correction factor for
drag flow and that for pressure back flow.

That relation can

be derived in two different ways.

The total flow rate and pressure distribution of the simplified
theory are compared in the first method with similar quantities
of the theory of end effects for the same operating conditions
J

Ua, va' q)'nd Po.

From the simplified theory

0,

==

Qt>o - Qpo

whereQPQ is proportional to Po.

The theory of end effects

predicts a total flow rate

Fig. 4.5 shows both Ql and

~2

as functions of Po for the same

Now FDE < 1, since QDl

reduces the drag flow (Fig." 4.), while FpE
1, since the
screw pump running at the same speed .

>

resistance of a channel with oblique ends _is smaller than that

of the channel with square ends, used in the simplified theory .
The two characteristics must then intersect.
The equation for thevelocity potential '(4.11) for the total
flow rate applies in both theories, but the sum of the conditions of Table 4.1 is in general not satisfied in the
simplified theory.

We know from the pressure distribution of

the simplified theory (1.).9)that the sum of these conditions

2

is only satisfied when ~ = - tan ~O (1.48).

"

The velo city

distribution for that case satisfies the equation for the

velocity potential, the differential equation (4:14) and the
sum of the conditions of Table 4.1.

4.1.9
The interset of Ql(PO) and Q2(P O) must then satisfy (1.48).
From (1.48, 4.43)

Q, ==

~ PC

+ +~(Po )

(4.4S)

while from (1.48, 4.44)

Q2. =
From Ql

= Q2

Q Do

( Ft>E

-+ FpE -l~4>o)

at that back flow coefficient then

\+

+0N\1.

'fa

Foe +

FpE.

~1.."'a

(4.f7)

This equation shows a simple relation bet ween the end effect
correction factors.

!!

Q, = TOTAL- FLow R.AE.

~I""'PL\FIED ~EO~'1

POINT W~t2e-

P->:_t~ceo

'foR.

(5)'2.. = loTL ~LOW t<A.TE: Fo12..

THEoe", o~ eN ~ E-F!=E:c..TS

FIG. 4.S - CI-fAeAC.R\s"\\CS foR. S'''-'PLIFIEt> Tl4E~"~

loJb Foe Tl-\EOI'<'1 OF

E-I-JI> EfFE.cTS

The same result can be obtained by the superposition of a

number of cases with different boundary conditions, which all
satisfy the condition tf /y = 0 along the flights y = + b O/2o
The flow rate in each of these cases is governed by the
distribution of the velocity potential along the inlet and
discharge ends.

4.20
First consider flow through the channel for uniform pressures

Po along the ends.

The flow ra t c: is QPE; the velocity

potential along the ends is given by (4.26).

The uniform

velocity potential at the discharge end is then

'fJPI

The pressure at x = a, t = 0, is equal to PO; that at x = 0,

y

0, is zero .

Next consider the velocity potential for the pressure back

flow rate of the simplified theory where again p =
x = a, y

0, and p

at x = y

0.

Po

at

The pressure gradient

in the channel direction is

The average velocity

o for this case becomes from (4.9)

(4/S"0)
while the velocity potential is

11
P2,

= _ hl Po

(~)

12,M

The flow rate for this case is equal to Qpo.

potentials

'I

x = a, y

0.

I
~Pl
j

The velocity

j "
and I,~P2
have the same magnitudes at the point

The difference between the flow rates QPE and

Qpo is from (4.36)

~IS2.)

4.21
This difference, QPE - QpO' is due to the difference between

lJI' Pl

and

6lf1p =

lfJpl

the potentials

l.f "p2

alo'ng the discharge end.

Then

//

- lfp'l - -

Substitution of {40l6} yields

~t.r p -

h'lpQ ( :a
12JA 0.... \ -t0,lV\.!(J

<)

The drag flow rate QDl was from {4023} due to a velocity
potential at the discharge end

That flow rate is also

The velocity potentials lllpp and

lfJ D are

both proportional to

y o The flow rates (4.52, 4056) due to these potentials must

then be related by

(FpE -

\) Qpo ::

(fDE - I) Qt>o
Substitution of (4054, 4055) and equatioris (lo29, 1030) for
QDO and QpO' respectively, yields the relation

=
or

Fl:)~ + Ffe -t~Ceo

+ ~(Po

This result is identical to (4047)0

This relation is shown irl

Figo 406 for a number of helix angles.

angles are of the order of 17 or tan

Commonly used helix

lfo ~ 1/3 ~

Figo 4.6 then

shows that the ends must have a larger effect on pressure

4.22

2.0

L<6

I.Go

(
..

'"

1.4
Fl)i;

\.2

r---I-.

\ .0

.-

""

----- r---- t--- ~

~ ts======t:

!'"

O~

. i""tcw\"'fo -

O.G.

21;_1' - J.

~~'eo~1

014-

0.'2.
.2.

.4-

F-\ 6. 4. ( -

.G,

.~

I. (;)

1',2

I. "f-

~
I.~

I. "

R~LATI~ BE:TwEE,J rpE I\JD F1:)E

Fot<..

D\~F~r<E-N

I-IE\..\X Al-JG;\..E-s>

2.0

4.23
back flow, while the end correction factor for drag flow must
be close to unityo
409

Calculation Re sult 5

The finite difference calculation of end effect correction

factors is described in Chapter 60

The integrals (4035, 4.41)

are calculated with that method from Laplace's equation and

the boundary conditionso

The accuracy of the numerical method

was ascertained for "a similar problem with a known analytical

solutiono

The accuracy of the calculated end effect factors

was further tested by varying the mesh size.

Both FDE and

FpE were calculated to investigate the agreement with the

relation (4047) between these factors and~oo

The numeri cal method breaks down when the number of meshes in
the y-direction becomes too largeo

The width bO must then be

divided into fewer meshes when the factors are calculated for

large a/bo ratioso

End correction factors can either be calculated as functions
of the L/D ratio of the screw, or as functions of the channel
ratio LCO/boo

The channel ratio is here defined as the ratio

of the developed channel length LCO (= 2 a) to the channel

width boo The following relation exists between L/D and
Lco/bo for a single flighted screw with a small land width

L l,tNv'fo

" TI' D sVvcteo

=lD) 'TT S~'fo

4.24
Tab1e 4.2 shows ca1cu1ation resu1ts ror FpE ror dirrerent helix
ang1es as function of the channe1 ratio LcO/bO.
width was here divided into n equa1 meshes.
for different n va1ues.

The channe1

Resu1ts are shown

Sma11 helix ang1es require a 1arger

number of meshes:' in the y-direction.

The factors cou1d, for

sma11 helix ang1es, on1y be ca1cu1ated for a channe1 width

divided into a sma11 number or equa1 meshes, since the required
number of meshes in the y-direction wou1d otherwise become too
large

The tab1e- shows an insignificant increase in accuracy

for n 1arger than 6.

Tab1e 4.3 shows simi1ar resu1ts for the drag rIowend correction ractor FDEo

The end errect correction factors are shown.

in Figs. 4.7 and 4.8 as function of LCO/bOand in Fig. '4.9 a1so

as function or L/D.

The ca1cu1ated va1ues agree withthe

re1ation (4.47) between FDE' FpE and the helix ang1e, as shown

in Tab1e 4.4
TABLE 4.4

Cf 0

6
6
6
6
6
6
6
6

tan
1
1

1/3
1/3

LCO/bO
3.0
4.0
3.0
4.0
6.0
4.0
5.0
6.0

FDE

FpE

0.8141
0.8667
0.7974
0.8724
0.9273
0.8496
0.9046
0.9304

1.1858
1.1332
1.8101
1.5101
1.3704
2.3530
1.8582
1.6265

1+tan2~o
2
2
Ik
Ik

1~

1-1/9
1-1/9
1-1/9

2
FDE+FpE tan <,0 0
1.9999
1.9999
1.2499
1.2499
1.2499
1.1110
1.1110
1.1111

TA8LE

2
4
6
8

10

L Ib:2
Co

2.5

1.3143
1.3019
1.2951 1.225
1.2910

N ~o lbo::: 2

2.5

2
4
6

8
10

2.083

l/r--:.3.5
Co

b""

4.0

4
6
8

2.399
2.3530
2.8030 2.334

Lel/ bo :

2
4
6

5.0
2.967

3.3177

I\.)
~

TABLE 4.3-DRAGFLOW END CORRECT ION fACTORS

TAN
N

LCo lt=2
0

2.5

3.0

3.5

0.7717
0.8048
0.8141
0.8183

4 ' 0.6856
6 0.6980
8 0.7048
10 0.7088

4.0

e=

1.0

4.5

5.0

0.8377
0.-8602
0.8667

6.0

7.0

0.8742
0.9132
0.8912 0.9109 0.9246
0.8961 0.9149
0.8979

8.0
0~9346

9.0

10.0

11.0

0.9465
0.9338
0.9423 0.9484 0.9519

JAN ~= 1/2
N

LCoIb~2
0

2
4
6
8
10

2.5

3.0

3.5

4.0

4.5

0.8510
0.8674
0.8724
0.8748

0.7899
0.7974
0.7292 0.8016
0.8041

2
4
6
8

LCo 1'6,,3.5
0

4.0

4.5

5.0

5-1/3

r:

5-2/3

Lco/bo=

5.0

5.5

6.0
0.9146

6.0

6-1/3

6-2/3

0.9219

6 .. 5

8.0

9.0

10.0

11.0

0.9560
0.9597

7.0

8.0

9.0

10.0

0.9392 .
0.9578
0.9438
0.9699
0.9345 1.0001
0.9951

0.9304
JANf~

7.0 .

1/3

0.8912
0.9014
0.9046
0.9424

0.8444
0.8496
0.1996 0.8522

6.0

0.8933 0.9170
0.9425
0.9248
0.9073 0.9273 0.9409
0.9901
TAN

5.0

1/4

7.0 -

8.0
009522

9.0

10.0
0 .. 9668

1100

1200
0.,9746

13.0

1400
0.9194

4.27
FI G. 4.7 - PRESSURE BACK FLOW
END EFFECT FACTORS

FpE

2.'

r-----.-------.,-----,-----,----r-----r----r----,

2. 0

f-----\-f-------I-~r_+--___t-__\__+--_+_--_+_--__l

t.q

f----\-f-------I-----,--\j-_\__-___t--_+_-+--_+_--_+_----l

I.<t

f---~&------I----+-\\-_\__---+--_+-------''r+__r----='_+_---I

17

+----t--\\'\---f---_+---\'~;___--+---+_____'\,;___-+------1

IS

+------1f-----1I~_T_--+--_+--~~._____+--_+_--__l

, . ~ +------1f----I-----"rr+~-_'_'I''---_+--__+_'~--''''c:-_+_---1

\.7>

~-----I--_+---+-----=~~~-+----+----+--____1

I. 'l +---~~~-f--_,_:=-t--_+---+----F""""'_:::::__=::'-k::___--I

( CI4ANN(.\'" '-N\bTf-\

,'h.

'9

10

1>1'JIDED INTo

E.<S;>VALF1NI--rt= 'DIFFl::RE.NCES)

FIG. ~L8 - DRAG FLOW

END EFFECT FACTORS

FOE

\.0

...

.~

vK;.-~
...

/~

o.~

1)E

~~

10-

~
"'<'~

---

~V

Vk ....

:7
Vi:
Ij

.'

~~

1-~",'t.<.J..
.,.A:..~

"~4-

""4

I V~h"
;'

j
{l

fl~~"
I
I

I 1/

o.~

fh'~

I
- r. ba

~o

10

~1O~4

1.01
O.q

1 ~::~I_

1~-2 1

~7"-

///

0.1 1

1///'

F1)E

0.61

/
1'\-:: 8
L/1,:)

O.S'

FIG.4.9-DRAG FLOW END EFFECT FACTORS

(

C~ANNE..L

W\t)T\-\

O\'-J\DED

It--JTo 1fl'\.1 E<S;lVAL

FDE

ME S.~t.S) .

N
CC)

4.10 End Effect for Large Channel Ratios

Oblique ends cause a pressure ,distribution ne ar the channel
ends that is quite different from the one ' predicted by the
simplified theory.

One must expect smaller differences between

these two distributions at points in the channel that are

farther removed from these ends.
Consider, for example, flow through a long channel with
~.

oblique ends due to a pressure difference between inlet and

discharge (Fig. 4.10).

Assume that the pressure gradient

opjOy becomes insignificant at a distance 1*/2 from the inlet

and from the discharge.

The influence of the ends is then not

observed in the zone with length 11

the zone 11 in the x-direction only.

The pressure changes in

1et (~P/~x)l be the

uniform pressure gradient in zone 11 for a, flow rate Qo

Leo ----------------~~

L*

2.

I2-

L"ttlJl

-L!*
2

I
I

~
I
F\G.4.Io-

L2..

-~

OBL\QuE
END

I--

...1

I
CHANNEL.. W\TH

LARGE t?ATIO

Leo/bo

4.31
Consider now the pressure distribution in the inlet and
discharge zones with lengths L*/2 for a flow rate Qo

The

distributions in these zones are for a constant flow rate Q

independent of the length of zone Llo

The pressure distri-

butions in the inlet 'a nd discharge zones are then the same
even when Ll = 00
equal to L*o

The total channel length for Ll = 0 is

Let FpE * be the end correction factor for a

total channel length L*o

A pressure drop

L~(~)

'Ox I

would yield a flow rate Q FpE * in a channel with oblique ends

and a total channel length L* , according to the definition
for FpEo

A flow rate Q would require a pressure drop

The pressure drop in zone Ll is for the same flow rate Q

The total pressure drop for a channel with a total length

LCO

L* + Ll is then

~Po= (~\

[L, + L*/F?tJ

Let FpE be the correction factor based on the full length LCO.
The total pressure drop is then also

4.32
From (4.62, 4.63) then

Substitution of Ll = Leo-L* yields

(~)
[I - .Fp-e.l J= .~to
bo

[\ -

~
r,, J

The right hand side of ,(4.65) is a constant if it can be

assumed that the influe'nce of the ends is insignificant at
distanc~s from the endslarger than L*/2.

This constant, e(fo )'

"

assumes different values for different helix angles.

Then

,;

for Le'o> L*

The parameter

e~

was

.from the calculated FpE

The results are shown in Table 4 04.

dete~ined

values listed in Table 4.2.

The parameter

e~

appears to be independent of the channel

ratio Leo/bo' except for small channel ratios.

It is, for

instance, constant for tan ~o = 1/3 and channel ratios larger

than 5.
Fig. 4.11 shows the change of
Leo/bo ratios.

e~

with helix angle for large

An almost straight line results when e~ is

plotted as function of cot

le 0

and

t.p 0

< 30

The data yield

theequation

Screw pumps generally have channel ratios Leo/bo in excess of

the critical value L*/b O and helix angles smaller than 30.

4.33

C'f

t
~

/~'-'NEAR.
/

APPRoX/fVlATION

'3
~

FIG .. 4.' 1-

CHA/l.jGE

0(

== Cat ,po

OF

C-e

\!VIiI;

Cet ~<:>

FoR. LARGE 04A"-''''L (('-'O<;

TA8lE 4.4-COEFFICIENT

cf

TANV:
1.0
0
N
2
4
6
8
10

Co

Ib=2
0

2.5

3.0

3.5

0.55 14
0.4783
0.4898
0.4638
0.4100
0.4557 0.4602 0.4611 0.4613
0.4508

4.0

4.5

LCo It~
0

2.5

3.0 .

3.5

6.0

7.0

8.0

9.0

10 .. 0

11.0

4.0

f=
0

1/2

4.5

5.0

6.0

7.0

8.0

9.0

10.0

11.0

0.4935
1.4953 1.4948 1.49471.4949 1.4945 1.4944
1.3698 .
1.3856
1.3862 1.3865 1.3861 1.3862 1.3858 1.3875
1.3429
1.3512
1.3514 1.3509
1.3000 1.3273 1.3317 1.3342 1.3193'
1.3180

4
6
8
10

TAN~o= 1/3

4.0

4 .. 5

5.0

2
4
6
8

2.4735
2.4747
2.3330 2.3470 2.3500
2.3510 .
2.3000
2.2770 2.3094 .23102 2.3111 2.3132
2.2513 2.2863

CG

t-;3.5

5-1/3

5-2/3

6.0

6-1/3

6-2/3

1.0

8.0

9.0

10.0

2.4146
2.4748
2.3510 2.3510 2.3506

TAN~ 1/4
N

l Co

tb0--

4
6

5 .0

5.5

6.0

6.5

7.0

8.0

9.0

10.0

3.4656 '3.4651 304650 3.4649

3 . 3150 3.3297 3.3331 3.3337 3.3339
3 . 251 0

I~

0.5583
0.558~ 0.5583 0.5585 0.5581 0.5580 0.5582
0.4903
0.49020.4904 0.4902 0.4902 0.4902 0.4902
0.4101
0.4702 0.4706
0.4611 0.4641

TAN
N

5.0

11.0

12.0
3.4651

13.0

14.0
3 . 4649

..

4.35
Equations (4866) and (4 . 67) can for those cases be combined
to express the end correction factor for pressure back flow
as a function of LCO/bO and ~o.

Then

(4.G8 )
As a function of L/D with (4061)

L/t>

Correction factors can with (4.68, 4069) be calculated for

any helix angle smaller than 300

The end correction factor

from (4069) is shown in Figo 4 . 12 as a function of the screw

to diameter ratioo

We now introduce the effective channel length Le' defined as

the length of a channel with square ends that has the same
resistance as t he channel with length LCO and with oblique
endso

Then

From (4068) for channels with L*

L~

,. . = 0.104
Po

I'

Leo
T
-

<

LCO and Cp 0
I'

09""

< 30

(.,

COLl, 10

The length L2 defined in Figo 4.10 is further

L2., ::. Lco - b o -Cat 'fo

The difference between the effective length and L2 can with
(4.71, 4 . 72) be expressed as

4.36

~
2J
UJ

-0-

U!.
~

l-

ti
0

L,

c,'

\.S>v

0
0-

r-.:.~

l,

rfJ

lil

IJ

UJ

IJ.-

ti-

LL

UJ

VJ

IA '
0

r-

IX

IJl

"3

C'I
0

.J

LL
~

\9

,"

en
I.U

C!.

:>
t!)

-1.Jl

VI

w
~
0-

.J fA

I:
~
0

<J-

"\j""
o

\9
LL

l.I'

4.37
The effective l engt h rat i o exceeds Lz/bO by an amount of the
order of 0075 for commonly used helix angle s when L*< LCO'
The drag f low end corr ection factor is close to 1 when L*< LCO
a nd helix angl es a re of t he order of 17 .
FpE = 1016 f or ~ O = 1706 and LID

5.

For instance,
The corresponding

drag flow corre ction fact or from (4.47) is 0.9a5o

The same

critica l l ength L* applies to the pressure back flow and drag

f low ca s e s i nce FpE and FDE are related by (4 . 47). The
numeri ca l val ue for L*/b O is discussed at the end of the next
se ctiono

4011

Pressure Di stribut i on and Streamlines

Cha nnel pressure di stribut i ons were calculated with the

numerica l met hod of Chapter 60 Fig. 4. 13 shows a typical
pressure distribution for pressure back flow for a channel
ratio Leo/bo

= 40

Pressures at all mesh points were calculated

for uni form pressure s along the channel ends and for p
the mid poi nt of the channelo

0 at

I sobar s were determined

graphically from that di s tributiono

Also shown are the streamlines o The s l ope of the streamlines
i s given by

For pressure back flow at any value of z from (4 . 7, 4. 8, 4.9,

4010)

The slope of the streamlines at any point (x,y) is independent

of zand is also equal to the slope of the streamlines of the
average velocities.

The isobars of Fig. 4.13 satisfy the

equation

The slope of the isobars is then

(~)

(4.77 )

\..d')( p -

The product

of the slope of the streamlines and that of the

isobars is then from (4.75, 4.77) at a point (x, y)

(

(d '
rJ~)l> c.(x.)s~

:.:.i:J;..
ct,.....

Isobars are thus perpendicular to streamlines.

For streamlines

from U... 74)

The continuity equation for this case can be expressed as

\r
-

'0'1

which shows that the expression vdx - udy is the exact differential of a function.fL , called the stream function.
lines are the lines of constant stream function.
iine

Then from .(4.78, 4.80)

Stream-

For such a

S=0.9 .

P"' IO

ir---t= f\

011- \0 II1

\ JJt~ I \ .....!L\:

\2

"v~

)....

I/', I

1=-----,--.. x

113
S=O
F IG- 4 '13 - ISOBARS AND STREAMLINES
fOR PRESSURE BACK FLOW

S= 1.0

Leo/bo = 4

/1~)(1
"\: . V

~ 2 -

/1

1'(
\ 1d~~
~

VELOelTY POTENTlAL

3/' "
17>~

'7

12

..,.
0J
([)

FIG-414-VELOCITY POTENTlAL DISTRIBUTION

AND STREAMLINES DUE TO Vo

The stream function is constant along the flights, since the

flights are streamlines.

(y

-bO/2) of Fig. 4.13.

valu x

Let~

= 0 along the lower flight

The stream function for a given

Xo

At x = 0
d.~

~
= -h

The stream function of (4.82) can be expressed as a fraction

of the stream function along the flighty

S ( '><.011 \)::

bO/2.

Then

~)
~/h

fl. (
-

'><0 I

The stram function for pressure back flow also can be expressed
with (4.9,

Then

This reduced stream function was determined by numeri cal

integration from the calculated pressure distribution.
lines are then lines of S = constant.

Stream

A graphical construction

was used to determine the streamlines from the S distributiono

Fig. 4.14 shows the distribution of the velocity potential
for drag flow due to VOalone.

4.41
For this case from (4.7, 4.8) for Ua = a

v-- _ (~~ );olr-hl+~(t) _

(~ ~ )r(~-h)
Partieles traveling at different z describe different streamlineso

For the average veloeities from (4012)

Streamlines for the average velocities are again perpendicular

to lines of constant velocity potentialo

A st re amfun ct ion ,

similar to the one used before, can be introducedo

constant along the flightso

It is also

Figo 4014 shows how only part of

the liquitl entering the channel from the oblique end passes
through the channel, while the other part leaves through the
same channel endo
The slopes of the streamlines at x = a (Fig. 40l3) change with
the channel ratio Lea/bo.

Streamlines at x

= a are almost

parallel to the flights when Lea exceeds the critical length

L*o

The pressure gradient ~ p/d x at x

in that caseo

( ~)
cl)<

= 0

is independent of y

The average slope of the streamlines at x = 0

lbo/. .

bo

r
~~) .t~
t J

:: -L
S (A\lG.)

0_

\..& x. 5

bo

_
--

_.,12.

~ d.~

t}e )
o(?>~

14,%~)
\.:

')(:::0

2..

was calculated by numeri cal integration for different ratios

Lea/bo.

Fig. 4015 shows the results for different helix angleso

4.42

,O.LO
I

-,

O.I<

O.I~

.
0.\4-

O. \'2..
-,

l~~)",~
Q.f

0.10

..

"bx,

A.T )(=0

~.o: eet tee

2.0

\.&
0.0<6

. ._-

\
\ \ \,
\

o.o~

0,04

>
~

C.ol.

2-

'3

~-

~~r-~ ~

FIG. 4.15"_ 'DECRf. A. SE

WITH

..

Lco

/10

o: (OP/'"'') A\JG .

AT CloJTEI (X=o)

p. JCeEASE

Lco/br-

O~

4.43
The average slope becomes insignificant when LCO exceeds L*.
We now introduce the length L** defined in Figo 4.10.

L*

L ~- "6 0 )t <fa

Then

( 4;q 0

From Fig. 4015 we se1ected va~ues L*/b O = 4, 5 for cot~o = 2, 3

respective1y>as the channe1 ratios for which the slopes are
insignificanto

Substitution of these ratios yie1ds a constant

ratio L**/b O = 20 The inf1uence of the ends then ceaseS to

be significant at a di stance from point P (Fig. 4010) equa1 to
b O when L**/b O

> 20

The critica1 1ength is then

5.1
CHAPTER 5
5.
5.1

Experimental Verification of End Effect

Introduction

Experiments were carried out to measure the end effect in

screw pump channels.

It was outside our scope to measure

end effects on rotating equipment.

The pressure distribution,

the end effect factor, and the streamlines were determined

for a stationary channel with a rectangular cross section
and an oblique end.
5.2

"

Model Channel

The general arrangement of the test model is shown in fig. 5.la.

A channel with a rectangular cross section is formed between
two clear plastic parallel plates which are separated by
two steel bars.

The channel is at one end connected to a

feed weIl, at the oblique end to a discharge weIl.

Corn syrup flows by gravity from a 2 inch diameter stand
pipe into the feed weIl.

The channel inlet cross section

is at the feed weIl perpendicular to the sides of the channel.

Fig. 5.lb shows a cross section through the feed weIl
and through the entrance zone of the channel.

A gate (fig. 5.lc)

is instalIed in the discharge spout of the discharge weIl.

The . gate is set to provide
oblique end.

submerged dischargefor the

5.2
The gap between the parallel plates is 0.125 inches.

The

channel is 6 inches wide and has an average length of 15 inches.

The aspect ratio of the rectangular cross section is 48/1.
The slope of the oblique end (1/3) corresponds to a helix

l..po= 18.5.

angle

Six small (0.032" diameter) injection

ports were drilled in the top plate at intermediate distances

of 1 inch.

These ports were in the first test used for injection

of small colored corn syrup streams.

Stand pipes were later cemented to the top plate of the channel.
Each 0.5" ID pipe was connected to the channel through a 1/8 l1
diameter hole.

Fig. 5.ld shows the layout of 17 stand pipes

in the oblique end and two stand pipes on top of the feed weIl.
5.3

Simultion of Screw Pump Channel

The model is an analog of the oblique end of a screw pump

channel with a large aspect ratio.
channel has a developed length LCO

The comparable screw pump

=

30" or a channel length

ratio LCO/bO = 5.
Fig. 5.2 shows the calculated pressure distribution in that
screw pump channel.

A similar calculated distribution for the

model channel is shown in fig. 5.3.

The pressures were in

both cases adjusted for a zero discharge pressure.

The

differences between the isobars for the model and for the
screw

p~p

channelare very smalle

The calculated end effect

factor ior the screw pump is 1.8582, that for the model
channel 1.8588.

Both factors were calculated by finite

A
SPOUT

I
I~-l---------B

I
1

I;- B

6"

LEVEL

-I

L1--Y-!j~

1.75"

-L

,I

I
I

-I

15"

I
I

.
_________ _ _______ _
L -irl. . . . L - - - - - - - - - - - -'--

CHANNEL
SECTION ICC:

,.

24"

- -t--- -

.. I

//

7/S"

bI-r-r-/ / / /

-- --+---

FI G. 5 .1A - PLAN VIEW

zzzz'1===t-~===

FIG.5.IC

.....,........,r-/. T

~-

C _ _~~
0.125

1\\\\~~

11

SECTION lAAI

1 r-r--

,-......... .

'

...........

FIG. 5.1 B

. . . r--

_
I

...... ~

-~

FIG.5.1

MODEL CHANNEL

- - _.....

_'-.

.-

_._'--'--

~.

I.-.lL1::::r::,.

FIG.5.1 D _ LAYOUT OF STAND PIPES

UI

VI

5.4
difference methods for a channel width divided into 6 meshes.
The extrapolated factor for the model channel is 1.836.
The small difference between model and ,screw pump channel
could have been predicted from the small relative transverse
pressure gradient shown in Fig. 4.15.
a screw pump channel with a length

The model simulates

equ'a'l:t<!> ~ 'th.e \, rit: ial

length L* discussed in section 4.10.

It is, therefore, a

reasonable analog for the end of screw pump channels that have
lengths larger than the critical length.

5.4 Streamlines
Streamlines were made visible by the injection of small colored
streams through small ports in the inlet zone of the channel.
A main stream of clear corn syrup with a viscosity of 25 Poises
was poured into the 2" stand pipe.

Colored corn syrup with the

same viscosity was pumped from a blow case.

Fig. 5.4 shows

the model during this test.

We showed in Chapter
to isobars.

4 that streamlines must be perpendicular

This not only applies to the average streamlines

but also to the streamlines at any depth in the channel.

The

streamlines of Fig. 5.5 were photographed against a background

of 1" square meshes.

The streamlines are not influenced by the

magnitude of the flow rate since inertia forces are insignificant

at the flow rates used.

STREAMLINES PASS THROUGH

THE 6

I I

I'

1\

I
2

~
-+-

,I

I /

psllO

--tt-1 ~"\JXJ ~

10

11

12

13

14

15

16

18

17

19

20

21

22

23

24

2'5

FIG-52 - PRESSURE DISTRIBUTION AND STREAMLINES

IN SCREW PUMP CHANNEL

_..._. ........................ THEORETICAL

~-U-~-I-+-I-1IW\H\rr~,,~\) ~~ ---

-1

!
!I---H--t--t-t'--

\
-t-

"--,

~-~~

~-

INJECTI ON PORTS

p =1

-e

. ,. . .

~~ ~
.
..... \'"

r-O .5 VI
,1
______

, ,
-~
~b " K
"
1
.' ..........
,,- . 1 5 [i .:x/.......
4,

~n

2,,~

EXPERIMENTAL STREAMLINES

,,// V _______ ______

vKr\ \ )~
\ \ j~; \

L-I-~--+-t--4-+-t-+-r-rI~~

STREAMLINES

I---r-------I----

~
5

FIG

53 - CALCULATED ISOBARS , EXPERIMENTAL AND THEORETICAL

STREAMLINES FOR MODEL CHANNEL

(]1
(]1

5.6
The photograph shows a widening of the colored streams towards
the discharge end.

This is caused by a reduction of the velocities

and not by diffusion.

Theoretical streamlines were calculated for the screw pump
channel (Fig. 5.2) and for the model channel from the pressure
distributions.

The solid lines in Fig. 5.3 are the observed,

the dotted lines the theoretical streamlines.

Only streamlines

passing through the location of the injection points are shown

in both figures.

The flow rate between adjacent streamlines is

1/6th of the total flow rate, except in the two outside lanes,
where the flow rate between the streamline and the flight is
1/12th of the flow rate.

The small differences between observed

and theoretical streamlines near the discharge end could be

due to inaccuracies in the photographic technique and due to
the difficulty to measure streamlines precisely in the region
where the colored streams fan out.
5.5

Pressure Distribution

Stand pipes and a separate 1/2" inlet connection to the main

stand pipe were instalIed for the pressure distribution test.
Colored corn syrup with a viscosity of 8.9 Poises was pumped
from a blow case into the main stand pipe.

The level in that

pipe was manually controlled at 12" + 1/8" above the top

plate of the channel by regulating the air pressure in the
blow case.

The level in the discharge weIl was kept at 1-3/4"

above the top plate.

Fig. 5.6 shows the resulting pressure

distribution against a background of I" square meshes.

5.7

Fig. 5.4

Model Channel During Injection Test

Fig. 5.5

Experimental Streamlines

5.8
The levels were measured in all stand pipes af ter an equilibrium.
condition had 'geen reached.

These measured pressures are compared

with the calculated pressures in Fig. 5.7.

were the

r~sult

The calculated pressures

of a finite differencecalculation for a channel

width divided into 6 equal meshes.

The measured pressures were

reduced in Fig. 5.7 to give a pressure difference of 10" between

inlet and discharge.

They show good agreement between theory

and experiment.
5.6

Experimental End Effect Factor

The experimental end effect factor, calculated from the

flow rate, the pressure drop, the viscosity, and the dimensions
of the model, was 1.88.

This factor is about 2-1/2% larger than

7..

the theoretical value.

The ineasured flow rate was ll.5 m 3 / sec.

It corresponds for a

viscosity of 8.9 Poises and a density of 1.34 gram/cc to a

Reynolds number of Re

0.15.

The length for the pressure drop

per unit length to reach a constant value is for the inlet zones
of circular tubes
z

0.0288 D Re

where D is the diameter of the tube.

Hartnett et.al.(23)

determined the constant in this equation for rectangular tubes

with aspect ratios of 10:1, 5:1, and 1:1.
then b replaced by the hydraulic diameter.

The diameter D must

Hydraulic diameters :C".:1

are for these aspect ratios respectively 1.81 h, 1.66 h, and h,

where h is the short side of the rectangIe.

5.9

Fig. 5.6

Pressure Distribution Test

o MEASURED

PRESSURES

10
I

t4

I
I

I
I r--INLE

DISCHARGE

1/

I
I

3 4 5
-L.

FIG. 5.7 - COMPARISON OF THEORETICAL AND

MEASURED PRESSURE DISTRIBUTIONS

5.10
Hahnemann and Ehret (21) investigated a channel with an aspect
ratio 50:1.

For our purpose, it is more convenient to use the

constant in the

~ollowing

equation
=

C h Re

C. h

(9 V'D H ~)

7'

The resul~ing values o~ the 'constants are 0.0125, 0.060, 0.076,

-

......

' . I

.~ .

and 0.057 for the aspect ratios

respectively.
~or

o~

50:1, 10:1, 5:1, and 1:1,

The hydrodynamie entrance length is negligible

the low Reynolds number used in

o~r

,test.

the parallel plates was made to a toleranee

o~

could at most introduce an error of 2-1/2%.

stand pipes were measured with an accuracy
- --" p-o nding to a possible error in the end

The gap between

+ O.OOlfT which

The levels in the

o~

e~~ect

:tl/8fT, corresfactor

o~

', _The viscosity was measured in a Bro"okfield viscometer.

+1%
Pressures

in the model were kept low to avoid deflection of the plates.

The experiment al results appear to agree with the theory.

-.,..

6.1
CHAPTER 6-

6.

Numerical Solution of Partial Dif.ferential Eguations

6.1

Introduction

Calculation of the end effect factors of Chapter 4 requiresI the

solution of Laplace' s equation Jn a numb.e r of families of r:~egions.
I

For instance, all screw pump channels with the same helix angle '
l

belong to a family of regions with the same width and incr~asing "
length to width ratios.

Different families exist for different

i

helix angles.
A new numeri cal method was developed for the solution of linear
partial differential equati'cfns
tn families of regions
.....

Courant

{24,25} discussed some of the elements in his lecture notes

before electronic computers were used.

v.

He believed that the

method could only be used for symmetrical regions.

Other methods

require evaluation of the function throughout the region by

relaxation (11) or other iterative processes (12).
to a set of

finite~ifference

The solution

equations is in the described

method obtained in closed form.

Iteration is not required .

In Chapter 4, we showed how the calculation of end effect

~actors
l

is reduced to the evaluation of two dimensionless integrale

(4.34, 4.41) in a dimensionless region.

partial differential equations

~s .

The study'" of many :linear

aimed at the evaluation of

similar integrals . or similar geometrJc properties.

For

i~stance,

Timoshenko (10) "shows that the ' t orque M transm1tted by a bar with
a rectangular cross section is related to the angle of twist

ft ,

the modulus of rigidity G, and the dimensions ' ' and 'b' of the
~ cross

section by

M == -kt G B ~b

6.2

The dimensicnless ccefficient k1 is suph .a gecmetric prcperty

cr shape f~ctcr fcr Pcisscn's equatich in a rectangular regicn .
This shape factcr is a functicn cf the aspect ratio. bja.

The

ccnstructicn cf equaticns cf the type (6.1) can be predicted by

dimensicnal analys).s.

The shape factcr can then always be cast

in the fcrm cf a d:f.mensicnless integral fcr a regicn cr a

bcundary ~

Twisting of bars with rectangular crcss secticns cf different

aspect ratio. is a typical example cf a prcblem invclving a
I

family
cf regicns.
...

The .family ccnsists cf_ regicns with cne

fi~ed side 'a' and a systematicalli lncreasing ~idt h t b ' .

,"

, /..

;Thehew methcd is always shcrter than an iteraticn methcd when

a given partial differential equaticn must be sclved in
families cf regions.

This is particularly true when the

desired result is a gecmetric prcperty cr., shape factcr, rather

than the value of the functicn in the whcle regicn.
The principles cf the methcd are first discussed.

Nxt we

shcw the advantages when the methcd is applied tc families f

regi9ns.

..

A shcrt discussicn cn dimensicnal analysis and shape

factcrs is then given.

'"

..

.!."

The accuracy cf results cbtained with the new methcd is' asscc iated with the precisicn c.f the. calculaticn.

The accuracy is

. demcnstrated fcr a prcblerri with ' a kncwn anal.Y tical scluticn .,.
Slcw visccus flcw thrcugh

tube with a rectangular cross

"
~

secticn was selected tc deni"onstrate :the methcd and the influence

cf precisicn cn errcrs.

,.,
"

This is ,fQllcwedby a discussicn cf the

calculati.n cf .the end e.f.fect factcrs cf Chpter . 4.

... ".

6.3
6.2

Principles of Method

The method is a finite difference method which has many elements in common with existing numerical methods.

The region

to be investigated is first covered with a network of regularly

spaced straight lines, each line being parallel to one of the
coordinate axes.

The partial differential equation is at each

mesh point replaced by a finite difference equation.

The

resulting linear algebraic equations relate the value of the

unknown function at that mesh point to values at neighboring
points.

The solution of these linear algebraic equations

approaches the solution of the partial differential equation

when the size of the meshes are decreased.
Figure 6.1 shows a segment of a region divided into meshes.
The region has straight boundaries AA' and BB' which however is
by no means necessary.
description.

Square meshes are used to simplify the

Each mesh point is defined by indices i and j.

--.lh~

'I

,~

,/

J_

"I~+I

" .... ."

...

~-')i

BI'

",J

"14

.t..,d-.

'\

,/

,
l"

"
I

+/ . cf

,.

.... ,/

.... !/

1''\

1''''
....

h
-,,,,-

'"

,/

.....

f-

\.,/ .

1''',.F

...

"

Fig. 6.1 - Segment of Region

S' .

Assume for this discussion that Laplace's equation

\72.F =

(6.2)

must be solved for known boundary conditions

The values

F ~, . j,) are unknown at all interior points (i,j).

Either the

function or its derivative is known along the boundary.

finite difference equation of the form

(6.3)
can be written for the Laplacian at each mesh point.
Let the ranges for j and i be respectively j
-

= l,n,

= l,m.

The set F (i,j) can for every value of i be regarded as a

vector

F"",
Vectors

(..

tF ~ i-I and {F 1 i+ 1 at adjacent stations

i-I and i+l

Equation (6.3) can be rearranged to

can similarly be expressed.

One can then write a set of n- linear equations

f(LH)I) = - f( ~,O) +4"(C I') -F(L,2) - F(L_\) \ )
J

1=((+1)2.):- f(~,I) +- 4 fU,'l.) - F(C,~ ')

F(lt-\) 1)

=-

r-(~,'2.)

F(L-I) '-)

+ 4f(C,1)_ F(L',4) - F (l-'-I) 3 )

(6.6)

6.5

= 1,n does not a110w use of the e1ements F (i,O)

The range j

and F (i,n + 1), since on1y n unknown va1ues F (i,j) are considered.

These quantities are either equa1 to known boundary

.!

values of the function, or they are re1ated to the known :

derivatives at the boundary points j

1 and j

n.

In the

latter case.

F (L) 2.) - F ( L) 0')

2.h
F ( l) "vi 1- \ )

(~.7

F (l)~ - \')

'2.~

or
"

J=(L)O)

F(L)'l) -

)t:

'2.h('O'1\

F(l)~.H')-=- F(t)~-I)+Lh(;~)~
We wi11 here assume that derivates rather than functi.on values
Equat;i..o'ns (6.6) then become

are given.

F(l4-\) 7..)

= - FU) \)

+ 4 F( L)1.) - F( ~ I ~) - F(C- \ )2. )

F((H::3l)

~ 'F(()'l.)

+ 4'F((,1) - F(L, ~) - F( ~-I;~)

---F(lH\~-')

'

("9)

. .;

= -F(c,'V1-'L) + 4'F(LIY\-t)

ot'

- F ((I~) - F(c-I) 11-\)

6.6
This set of equations can conveniently be written 'in matrix form
... ..;
"-.

4 -2
4 -I

-I

~ F1 C+\;

'2-

[Fl- fF1, +

4 -1

\;..

o .

L-l

-14-1

((,.10)

- '2 4
On~ can

h~
b'1)'

.~

-'2.h~)~

with this matrix equation always ' calculate (F}i+l from

and the boundary conditions.

Consider now for

"

simplicitya case with a straight, boundary at i = 1 (Figure 6 . 2)

and a uniform value Fl along that boundary.

F\G.

<0.'2. -

U~K'NOWfoJ S

:. ,
,,'

INTRoDvc"Ojo.J o:
')(}

~T ,

= 2 ,'

l"

6.7

4 -2
-\ 4

Let further

-I

-I

4-1

-I 4
Equation (6.10) becomes fo;

-I

-2. 4

tF13

_1.l (~)~

The only unknowns in this equation are the elements F {2,j} in

vector {F1.2.

Introduce now the unknowns

Equation {6.ll} contains n-unknowns xk and a set of constant '

coefficients.

Formally we augment the vector {

element xn+l = l .to accommodate constants.

XI with the

Equation {6.ll}

then becomes

lr l, = 1Bj tE}~ [Xj

L-f\ + zh (~~),J

..

-- FI

-F,

Ol

'

- F,

..

{XJ

[-F, - 2.h(~\~J 1 -

(G. \1)

6.8

where

At i ==

.,

CD

.,

IA

[XJ

then
0

fFJ~ :

'"

10

.,

..
., ..
..

{B] tE1~ ()(1-fE1~~)(1 +

..

"
0

..

2 h (~f)1

"

~XJ

-'l~( uF)
().., .....

The recursion equation (6.10) is used to produce a series of

"matrices

tEl

boundary i
i

= m,

i .

We will assume that the region ends at the

m with a uniform derivative (~F/~X) = F'm.

At

then

1!_t::\_ F("rII\+ I) ~ ) X

- ' 'M -

~ ( ~_I \ ~ ~

"l.h

The matrix {E1m+1 must also be calcu1ated.

Equation (6.15) becomes in matrix form

Ct

~ E~~~\ - fE 1'"-1lX 1=

&

"

..

2hF~

..

2h F~

'lhF~
2~.f~
2hf~

{X1

(<0# \G )

6.9
All coefficients in this equation are known.

It can be

writ~en

as

wh'e re {G

is an n x (n+1) matrix.

This matrix equation represents a set of n-equations in

n-unknowns x k and a set of constants, since xn+l
unknowns can be solved by any convenient methode

1.

The

Back substitu-

tion of the resulting values for x k in

lF)~:: ~E]t

tX1

@,. IS)

yields the values of the function at any mesh point.

A rectangular boundary was used to simplify the description.
The method is not restricted to this simple geometry.
general region is shown in Figure 6.3.
intr:bduced at i

lan,d 2.

A more

Four unknowns, Xk, ' are,,

The reinaining unknownfunction v?-lues

can at i = 2 be expressed in x2' x ' and the boundary values,

3
using finite difference equations for irregular stars.

~----II---I

4
~"'~

Ir"nA\

Sr'Acr

6 .10
The unknown xl is eliminated from {E}6' using the boundary
condition at point P.

The remaining unknowns are evaluated

from the boundary cnditions at points Q2' Q3 and Q4.

The

method can in principle be applied to any shape of the contour

and any set of mixed boundary conditions.

6.3 , Extension to Families of Regions

The method is particularly suited to the solution of linear
partial differential equations in families of regions.

For

instance, consider the family of regions of Figure 6.4, characterized by a uniform width a and a variabIe length Lx.

Assume

that a given linear partial differential equation must be solved

for this family O'f regions for the same boundary conditions.
Let m be the mesh number corresponding to the maximum length

of that family. ,
Matrices of the type [EJi are foI' any length Lx the same in the
zone 0 ~ x ~ b.

A recursion equation of the type (6.10) applies

in the zone -x ~ 'b ; Another series of matrices of the type

can be formulated for the maximum range b

<x

Lm.

f Eli

The number

of matrices used for a given length Lx then depends on the

number of meshes in the x-direction contained in that length Lx.
A coefficient matrix, calculated at a given value of i, does not
change with the length Lx.

An in:creased length requires

additional matrices, but previously calculated matrices remain

unaltered.

6.11

..-t-1
/-

~F=ro

a..
\

~-

Lx
/~

~f=Fo
1

0.,

~ .......

t: C+o<. .

L"" '=

/I

/
..X
~

,\

1\
I

,I""--i

'2.

1.

IY\A'X. L.E:t-J G,l-l

6.12
The function was at the boundary x ="Lx assumed equa1 to a
given va1ue Fa for any 1ength Lx.

Let i = c be the mesh point

at that boundary for an arbitrary 1ength Lx.

The unknowns xk

then can be ca1cu1ated from

A simi1ar set of unknowns can be ca1cu1ated for a new region

terminating at i
desired va1ue.

c + ot , where the increment

0(

can have any

Then

This technique can be used to calcu1ate x k for a series of

aspect ratios Lx/b. Back sub~titution of any ~et xk in the
appropriate matrix equations ( i ' c) yie1ds the solution of the
differentia1" equation "for the required shape factor.
Solution of the differentia1 equation i,n an additiona1 region
requires eX. mtrix operations of the "type (6.14) and subsequent
back substitution of xk in the ap"p ropriate matrix

equation ~"

Additiona1 techniques are described in Section 6.9.

6~4

Dimensiona1 Ana1ysis and Shape Factors

Differentia1 "e quations are either solved to obtain the

qistri~

bution of a function in a region, to derive an integrated "form

O

of that function in that region, or to derive an integrated

form of the derivative on a boundary.
often the more importantpesu1t.

Theintegrated form is

6.13
The dimensional construct ion of equations for integrated
quantities can always be predicted by dimensional analysis of
the differential equation.

Exact equations
for these quantities
...

must always contain a factor expressing the influence of the

relative shape of the region.

These factors will further be

referred to as shape factors.

This will be illustrated with

the following example.

The axial velocity w is for slow viscous flow through a long
tube of constant cross section governed by Poisson's equation

'n~_

,~h ,
-;tA-dt

C
IC

where the pressure gradient is a constant.

Geometrically similar tubes are characterized by a similar
dimension 'a' of the cross section. The cross sectional area
is then proportional to a 2 Introduce now

- =

lAr

t..V
o..,'l.. d P
)J.. d~

_.

:>

'><
C\..

'<t

:.

(;,.20 )

0,..

Equation (6.19) then reduces to

o'-~
'0')( ....

O"-::::S

+ -0'1'-

((,.2/)

:::;

Tpisequation is identical for all geometrically similar tubes.

)

The boundary condition

W =

.).t, . th tube wall applies for all .

tubes . . Tb.esQlution w {x, : y} of {6 .21}

.~

sat;isfyngthe

':boundary conditioilS'" :is ttben the same f'r ' geometrieall,

:simi'l a-r t llbe s ,

6.14
The flow rate through any of these tubes can be written as

The integral of (6.22) has the same value k for geometrically

similar tubes.

The flow rate can then be written as

where k is a geometrie or shape factor.

Next we investigate families of regions , withthe same characteristic dimension

'?'

in one direction, but having a different

characteristic dimension in the other direction.

consider 'the family of regions of Fig. 6.4.

For instance,

All regions have

the same : height b but the length Lx is variabIe.

The flow

rates through tubes with these cross sections can always be

expressed as (6.23).
different ratios Lx/b.

The geometric factor k is different for

It is then a function of.that ratio.

Other important integral parameters appear in equations for

flow or fluxes across boundaries.

Integral constants ofthat

type were used in the calculation of the end effect factors in

screw pump channels (Chapter 4).
The matrix method can be used to calculate the shape factors
f~Qm

the ' coefficient matrices without having to determine the

, distribution of the function itself.

6.15
Calculation of the flow rate (6.22) . from the matrices [Eli is
here shown as an example.

The flow rate through a square with

s ides 2h with point (i,j) as center is by Simpson's rule

+h
,6.Q,=

+-~

Jf

w-dx d'}

<'"

- h -h
~

~~ [I~W-(~)~)+ 4lVj(L,-I)+W-(~I~+I)+-~((+\)d)+w{l-',d)J
.
+ LJ(C-I,&-I) + w(C-',~+-I)+-~l(+ ,)J-I) + W(C+I,J 'H)]

where

I
Q.lt =

/C:,

4
The flow rate through the strip between i - 1 and i + 1 becomes
for j

l,n

661 0 =

hl.
-9

C\~~ wkQ.

-+e.

where now

(I)

4-

a.~Q

4-

I~

2.

2.

IC,

'2..

"
<I-

~t

-7-

4
(~)

(&>. 2()

6.16
The veloeities at i contribute a quantity

Qi to the flow

rate, given by

In analogy to equation (6.18), let

[wl~:tt1~{x]
then

6Q~ ~ [~t

tEL {XJ

Similar expressions can be written for the ' 'contribution to the

'f low rate from

[wJ i-~

fvv} i +1

and

The summation of the flow

rates through all strips in a region yields an equation for the

total flow rate

f F\

where

is a row vector.

Equation (6.30) shows that flow rates and shape fa.c tors are
linear equations in xk.

They can be calculated once

{x} is

known.

There is "no need to calculate the velocity distribution

by back

substitutio~

if the flow rate or .shape factor is the

only quantity of interest.

6.5

Accuracy of Numerical Method

'.,

The principles of the numerical method are

s~raightforward.
,

The method is clearly not attracti ve for manual cp.lcilltion of

aproblem.

The limitations to the method are demonstrated" in

the following sections.

The key to success lies in the pre-

cision or number of digits retained in all operations.

"

"

6.17 The accuracy was tested for the problem of slow viscous flow
through a long tube with a rectangular cross section, for which
the analytical solution is known.
in the next section.

This example is discussed

In addition, we calculated flow rates

.
,
through triangular, circular and semicircular cross sect10ns,
for which shape factors are known.

6. 6 Shape Factors for Tubes with Rectangular Cross Sections

Slow viscous flow of an incompressible liquid through a tube
with a rectangular cross section has been studied by many
investigators.

The differential equation for viscous flow is

analogous to the equation for torsion of bars with rectangular

cross sections.

The resulting flow equations are further used

in some of the screw pump equations.

For instanee, Squires (4)

uses the equation for the pressure back flow rate in the form

for a channel with sides 'a' and i tb' . The quantit:y Q~

is here the flow rate between parallel plates with a gap 'a',
a plate width tb', and the same pressure gradient.

The so-

called shape factor Fp is a function of the aspect ratio bja.

The difference between (6.23) and 6.31) is a result of a
different definition for the geometrie constant.
Fp are functions of the aspect ratio,since

Both k and

6.18
Flow through the tube is governed by Poisson's equation (6.19)
for w

= 0 at the tube wall.

The analytical solution yields the

shape factor of (1.37)

Figure 6.5 shows the cross section of a typical rectangular
tube covered with a network of lines parallel to x- and y-axes.
It is only necessary to investigate one quadrant of the cross
section since the solution in the remainder can be found from
symmetry.

The region has m-meshes in the x-direction, n-meshes

in the y-direction.
indices i and j.

Interior mesh points are again defined by

Boundary conditions are

vv-( 0) ~ )
W-(X)o)

)
oX (b
2)~

'Ow

.....
':

c::>

(Co. ::'3)

= 0

oW- ()( B:.) =

0'1

'2..

Finite difference equations become for this case

vs(t:.-I-\)\ :

4w(C)I) _w-(~)2.) -!J(e-I)I) +h'-c~

IN ((.+1)"-.) ': _W-(~)I)+ 4- w(C)2.) _w-(~);,) _ W-(~_I)'2.)

+ h'LCo

----------- - --

The vector

where

tW(i for the velocities at

i can be expressed as

6.19

11

",3
[>

o
11

~Ix

(O~

~-~-- --?-,

al

u..o

o
~

:r

'2

VI
l!J

0
l-

c.)

I
~

....,9

\9

- ,

u..

(l

v1
0

0!.
(J

6.20
Matrices

fE1 i

(6.10).

Initial matrices at i = 1 and i = 2 are

can be calculated with an equation similar to

" .. ..
. .. ..
Cl

lEl, =

Cl

al

Cl

f01 ) tE11-=

G-

<t

m + 1.

The center line of the cross section is at. i

The

derivative ow/vx vanishes along that line.

The unknowns x

are then calculated from the matrix equation

The flow rate equation (6.26) must be modified, since this

region contains only half.a strip.

The velocity further

vanishes at the boundary, while the other quadrants of thecross section must be accounted for.
result for n = even and n

odd.

The flow coefficients are

!Y) :: E\J Eo""

fV\ :: 01>1)

3'2. ~
\'2. <a

"17.

ICo ~4

I"
'32

3'1.

32- Il~

Different coefficients

~'2.

~e.=

32.

ca

,,,

I 'l. ~ 32.
~1

IG

~2.

I"L~

32.

~e=-

I"

'4

IC,

32 1'2.<b 32
<0 3'l. ~

(~:~'1)

32 12.<!. 2(~)

\ Co

(:,4

I'

\~

64

\~

('V\.)

-.

'

6.21
Let now

fK~

(~112, J~, -----[kJ = {S,32, I", -- --=

I", ~2.1 ~ ']

32,

~Q :
For i

= odd

{k

for

tG)!>2, IG, 16J

= even

The incremental flow rate for i

- even
(6.40)
n - odd

for n

is then

1[E L- + 2 E~1tX1

(6.41)

it is
- (6.42)

These alternating equations apply only when the derivative in

the marching direction vanishes at the center.

The flow rate

Q~~ of (6.31) is
p

&:
I

=:

C~ 1"\3-6 _ 1..
12.

VI-

c h4'"v\.3~

(6.43)

There is no loss in generality when we assume Co = 1, h

=1

in

(6.43) and in the finite difference solution of the problem.

The shape factor is then calculated from (6.43) and the sum Q
of the incremental flow rates (6.42).

This sum can be expressed

as
(6.43a)
The shape factor is then

Fp

= ~; - 4~~

(6.44)

6.7 Errors and Precision

The accuracy of the numerical solution is influenced by the mesh
size hand by the precision of the numerical calculations

Both

the finite difference equation for Poisson's equation (6.3) and

Simpson's rule have errors of the order of h4. Derivatives have
errors of the order of h 2 Shape fa'c tors for viscous flow
through rectangular cross sections then have errors of the order
of h 2

6.22

The mesh size h is for a given width inversely ,proportional to

the nurnber of meshes.

The exact

shap .~

factor Fp canbe expressed

in the calculated factor FpN' the nurnber

error coefficient

t..

o~

meshes n, and an

Then

For two different nurnbers of meshes nl and n2

Substitution leads to the extrapolation equation

Extrapolated and calculated shape factors are shown in Table 6 01.

Exact shape faotors were calculated from the analytical sol uti on
(1.37) with errors smaller than 10: 8
The last two lines of Table 6 . 1 show the exact and extrapolated
shape factors .

The difference between exact andextrapol a t ed

= 6, n2 = 8 of the order of 00001% 0 I t i s of

the order of 0.005% when extrapolation must be based on nl = 2 ,

factors is for nl

n2 = 4.

Shape factors for other contours may require more com-

plioated extrapolation equations .

Calculation resul ts can then

be refined wit h .,a . least square fit .. Such a technique reduces

the need for calculations with large numbers of meshes e
Inaccuracy can further be

cause~by

:."1'1'

and by ill-conditioned matrices '.

...

accumulated round-off errors

Errors from both these sources

~.

23

TABlE 6.1-INFlUENCE OF MESH SIZE ON SHAPE FACTORS

ALL ,FACTORS CAlCUlATEO WITH 16-PRECISION

NO.OF
MESHES

1.000

0.666

ASPECT RATIO
0.400
0.500

0.333

0.250'

N = 2

.39843150

.. 5664212

.66929251

.73440324 ' ,. 71847639

N = 4

.41590522

.58214633

.68,189031

.74470441

.18111066 , .84030712

.41914600

.585004626 .68420329

.74659291

.78855752

~90505584-

.42027823

.58600208

.68500912 - .74601236

EXACT ....

.,42173108

.,58728215

068604505

.74809524 ' .18995080 ' .84243888

EXTRAPOlATEO

.42173395

.58728499

.68604158

.. 14810182

.18:998875

.83381151

.84247030

-INSUFfICIENT PRECISION
-.CAlCUlATEO FROMANALYTICAl SOLUTION

TA8lE 6.2-FLOWTHROUGH TUBe WIIH RECTANGUlAR l CROSS-SECTION

INFLUENCE OF ,PRECISION ON SOLUTION OF LINEAR EQUATIONS .. )
ASPECT
RATIO

MI

0.50
0075
1.00
1.25
1.50
1.75
2000
2.25
2.50
2.15
3 .. 00
3025
3.50

2
3
4
5
6
7
8
9
10
11
12
13
14

--)

Xl(8-PRECISION)
19.598685
102 .. 14011
113.78681
119051310
122.23602
123.50958
124.15718
124.30282
124.34753
125.88144
119.43261
110.,81393
-15.60829

X2(16-PRECISION)
19.598101
102 .. 140115
113.186165
119.512418
122.23680
123.5.1115
124.10229
124.31543
124.50140
124.55945
124.58619
124.59851
124.60418

.. ) BASO ON ~ QUADRANT WITH N=4 UNKNOWNS.

= NUMBER OF MESHES ' IN MARCHING DIRECTION

-*) MI

RElATIVE ERROR
-0.-000,000,039
0.000,-000, 81:>
0.000,005 t ,)
0.000,-006 , 3
-0.000,013
0.000,44
-0.'000,58
-0.001,2
0.011
-0.041
-0.11
-1.12

6.24
are reduced when a higher precision is used.
defined as it is used in digital computers.
,

Precision is here
A number can be

expressed as + mp x 10:-=- where the mantissa mp is the largest

decimal fraction smaller than 1, while q is an integere

The

precision p is the number of digits retained in the mantissao

The round"':Off error of the mantissa is of the order of lO-(p+l)~
'r,he relati ve round-off error ofthe number is of the same order"
The accumulated round-off error increases with the number of
elementary operations..
and m.

For

instanc~,

That number increases rapidly with n

the number of operations required for the

solution of n linear equations in n unknownsapproaches n J /3 when

n is ,largee

The number of operations for 'the calculation of

E1t is proportional to the product n m.

The growtp of the

error can become p'r hibi tive' when precision is insufficient .. '
The magnitude of errors was investigated for the problem of
viscous flow through the rectangular tube by comparingcalculatio.n results for different pr'ecisions p

4, g, and 16.

Round-

off errors can occur or can be propagated in the calculation of

f E1i'

inthe' solution of the ~'~t of linear eqllations, and the'

subsequent back substitution of xk in flow rate equations.

The series '{Eli starts with two matrices with binary coefficients.
that have no round-off errors.

The same holds true for the

finite difference equation or for the matrix

[B}"

The coeffi-

cients of subsequent matrices have no round-off errors until

one f the elemnts exceeds lOP.

The typical increase of 'the

6.25

)
20

l A.C2G E;c:;-r
ABSoLUTE
COEFF\CI~NT

..... --

(~::.\>~\
I(

15'

A~S.OLVTE
CO~FFICIt;IJT

<..:f<.:: I) 'Y\..)
IQ

'a_ PR.~C. \S IO"J

---1-

4- PREC.I ":,101-.)

20

IS

10

30

- -...
-

FIG.~:

(:, -

OF- ,\-IE

\..;

fV\~Tel')(

IN"CREA.SE

"JVMBe.

Or Tl--\E- A&SOLVT'i:-

COE.~FICIE~S

I/'J

tE}r.

W\T\..I

VALvES

6.26
coefficients of fE}i is shown in Figure 6.6.

Here we plotted

the maximum and minimum absolute values of the coefficients for

k

= 1,

n as function of the matrix number i.

The boundary

matrices ha"{e no round-off errors when the corresponding { E}

matrices have no errorsG
These boundary matrices . were solved with 8 and 16 precision to
determine to wht degr~e they are ill conditioned.
are precisely identical for m

< 13

Th.e matrices

as shown in Figure 606 0

The

solution (xk)$ for 8 precision differs from (xk)16 for p = 16"

Matrices are more ill conditioned when the relative
between thesesolutions is larger.
error for xl and n

differen~

Table 6.2 shows the relative

= 4 forincreasing values of m.

The coefficients ~F} of the flow rate equation (6 . 43a) showan

increase similar to that f {E}i.
~agnified

The errors 1n xk are then

in the flow rate equatione

The flow rate coefficients

and the relative error in xk increase simultaneously.

The cal-

culation of shape factors quite suddenly becomes unstable as

shown in Figure 6.7.

The point of breakdown is very much inf lu-

, enced by the precisione

<,/"

The influene of the number of unknowns n was for ' the same
problem investigated for a uniform precision p = 160 The results
are shown in Figure 6.8, indicating that the problem becomes
unstable . 'at the same value m = 20 for any value of n o
The

ill ~ condition

of matrices is asociated with'the distribu-

tien'of the coefficients.

The character of the square pa:nt of

the matrices can be described easier for flow through the full

6.27

1.00 '

"

,
\

..
Jt

'b-PRr=:CIS10N

.50

~I\

.\

-.

\
\

.\

A~PE:c;:r

"b.,IO
-

----.

4-

INFLVENC~ O

SHAPE-

2.

1=\G.6.7-

\
\

' PRECISIO~ 01=

FACTOR..

CA.LVVL.o.TION

ST.t\BII...I''1

Fol.

'Yl=

. 1.0

Fp
SI-4APE
FACToe.

O'Jr:;-

0,5"0

m..

No. OF ""ES4~S II-J

(V\A,~C\.t It-JG DI~E-c..TIO'"

B -

1'2

'2..0

oF

4-

6.2e
cross section than through one quadrant.

T~e

number of unknowns

n is there larger and the square matrices for the unknowns are
symmetri.cal.

A typical set of matrices is shown in Table 603 0

The element.s on a line parallel to the principal diagonal are

the same, but the signs c,hange from line to line.

The series

of matrices starts with a .z ero matrix and an identity matrix o

f

The next matrices are the ~ri-diagOnal matrix

diagonal matrix ~B} ~I}, etc.
higher polynomial in

[Bl.

tB},

the five -

Every subsequent matrix is a

The distribution of the coefficients

is fully described by the distribution of the coefficients on

the secondary diagonal, as shown in Table 6.4.
There is no loss in generality when the coefficients of each
row are divided by the pivotal elements.

The resulting reduced

matrices have coefficients equal to 1.0 on the principal diag- .

onalo

The distribution of the absolute values of the reduced "

coefficients on the secondary diagonal is shown . in Figure 609.

The index s is here the distance of an element to its principal
diagonalo
Matrices of a series with n unknowns become more ill conditioned
when m is large.

All reduced coefficients increase with m.. '

Inversely, matrices for n unknowns are more ill conditioned when

the reduced coefficients are larger.
The calculation of shape factors becomes unstable at m
all values of n.

20 for

The distribution of the reduced coefficients

near the principal diagonal is from Figure 6.9 the same for any
value of n.

Inversely, one might say that the matrices become

6.29

TAStE 6.3
SERIES OF MATRICES fOR fLOW THROUGH
fULL RECTANGULAR CROSS SECTION

.,
1

tt .

...

...

...

...
...

...
...

...

...

...

tt

...

...

tt

...
...

tt

...

tt

tt

...
...

...
...

tt

tt

...

tt

...
...

1
tt

...

...

...

...

...

...

...

...

tt ,

tt

...

...

...

tt

...

tt

-8

17

-8

...

...

-1

tt

...

tt

-8

11

-8

...

...

-1

-1

...

...

tt

-8

11

-8

...

tt

-1

- 1

...

...

-8

17

-8

tt

...

-1

- 1

...

...

...

-8

11

-8

...

tt

- 1

-8

17

-1

tt

12

-1

...

tt

80

-49

12

-1

...

...

- 4980

-49

12

- 1

401 -280
-280

97

-16

401 -280

97

-16

..

401 -280

91

-16

401 -280

97

-1 6

401 -2 8 0

97

97 -280

97 -280

-1

12

-49

80

-49

12

- 1

-16

tt

-1

12

-49

80

-49

12

-16

-1

12

-49

80

-49

-16

12

-49

80

...

...

1=4

-49

...

...

tt

..

...

- 1

...

...

-8

12

11

- 1

tt

...

...
...

tt

tt

tt

...

-1

...

...

tt

...
...

...

1=2

1=1

tt

...

...

...

-1 .

1=5

91 -280

91 -280

- 16

( =6

401 - 2 8 0

97 -280

40 1

...

6.30

1.0

Ja.o::----....-------r------,----~---___r

2S =

rv.ATRIl(

No.

t
o

\
-.-..,...... S:

F'G,~." -,J) IS TRI e.uTIOI-J

ON

TABLE
MATR IX

t!Q.

6.4

'3

'2...

O~ R~J)vCE-D

SE C.ONCA~

ELE"'1ENTS

D,o..GON~\..

COEFFICIENTS ON SECO.NDARY

.5'

l)\STA"-'C,E RGoM PRt\JClf'I.. DlA.<;OI\JA\,...

DIAGONAL

2
3
4
5
6
7

4
17 . .
80
'

-I

-8

~49

l
12

-I
40 I -280
97 -IG
1
2084 -1569 G72 -161 20 -I

6.31
ill conditioned when the reduced coefficients on the first line
parallel to the principal diagonal exceed 0 .. 920
The instability is here discussed in terms of the instability
of shape factorso
calculationo

The errors in xk are magnified in the flow

This is not so forthe calculation of end effect

factors where instability for p = 16 occurs at i = 30 for any

..

number of unknownso
on the end useo

The limitations of the method then depend

For instance, calculation of the function in

I

a region can be very inacGurate while : the calculation of the

I

flux across a boundary is at the same time very precise.

culation of the velocity at i = 12, j

Cal-

= 4 in the quadrant of

the rectangular cross section requires by way of example substitution of the unknowns xk in the equation

where

')(., = \ .'24~
~'l..': I. q 1 S'"

)<.1.

'&

2.:' ""
4<1'

~4::' '2.,

The individual products are in this equation very much largr

than the velo city that must be calculated..

Accurate results

can only be obtained with the proper precision",

No rules can be given for the solution . of a general problemo
The values of the reduced determinants Dn are sometimes used
as anindex for the degree of ill conditioningo
The shape factor calculation becomes unstable when

6.32
where Dn is the reduced determinant of the boundary matrix.
That index is almost linear to m, as shown in Fig. 6.10.

Then

(6.49)
The required precision for m is for this problem approximately

A number of unknowns are introduced at i> 2 when similar

problems are solved in closed regions with convex boundaries.
The reduced coefficients are for the same n and m smaller than
those for the circumscribed rectangIe, which would tend to
make the problem more stabIe.
Round-off errors in {E}i are rather smalle

The relative error

is for two elements of [E} for the shape factor problem shown
in Fig. 6.11.

Here we show again the relative difference

between the elements calculated with 8 and 16 precision. These

errors were in this case calculated for boundary matrices that
have no round-off errors until one element exceeds 10P.
The matrix {E}l of the end effect calculation has round-off
errors.

All matrices {E}i for that problem then have round-

off errors.

The influence of round-off errors in { E} i was

investigated for the shape factor calculation.

That problem can also be sol ved using 0.1 {
of {

Ir.

11 for [E} 2 instead

12 and all subsequent matrices then have

The matrix { E

round-off errorswhen the calculation is performed on a binary

computer.

This is so because 0.1 is a rounded binary number.

6.33
IL.-----------,.----------.-----------r-----------.----------~

IOI----------~----------~--~------r---------~----~~~--~
~

'V\..-;.4

~ t-~~\+-------r----~~~-------+------~
#

2... .

FIG.

<0.

10-

10

1'2.

\4

C.~AI_JGE O~ -~~Ib'n.-I WIT~

"f'f\.

-7

Z ~ /O

EI-EI'I1E-AJT

e"

~ ""'Af2/)( NO,

1'2.._ _

~L_

__

_ _ _ __ _ _ _ _ _

20

~~

________

24

(..

__________

__________

3,

6.34
The matrices tE)i have from Figure 6.6 no
i

<23,

when based 'on p

16 and [ I}.

~ound-off errors for

Table 6.5 shows the differ -

ence in Fp for the two cases, indicating a minor influence of '

the round-off errors of { E } on the final re sult.
One can then conclude that instability of the calculation is
governed by the degree of ill conditioning of the matriceso
This degree will be different for other partial
equations.
equa~ion

differen~ial

are for the same partial differential

Matri~es

and the same boundary conditions, for n unknowns and

m steps less ill conditioned the smaller the number of interior

points in the region.

Integrated results for fluxes (4031 , \035),

which are based on derivatives, ca~ be calculated more accurately

than functions or integrals of those functions.

It should be

noted that the coefficients for the constant xn+l influence . the
stability of the solution.

For instanee, the end effect factor

(Section 6.8) for pressure back flow becomes unstable slightly

soonr than that for drag flow.
matrices are here the same.

The square parts of the

Different boundary conditions

yield different coefficients for xn+l for that caseo

6.8

Calculation of End Effect Factors

The end effect factors of Chapter 4 have to be calculated for

a number of families of regions.
regions wi th the same helix angle
ratios LcO/bO.

Each family consists of

~,
o

anq different channel

-

Helix angles are restricted to'those with mesh

points at the inters'ections of the oblique boundary wi th horizontal mesh

lines~

Half of the region of Figure 4.2 is in each

6.35
TASlE 6.5- INFlUENCE OF ROUNO OFf ERRORS IN E
ON SHAPE FACTORS fOR 16-PRECISION

MATRICES
10-~NKNOWNS

H/B

Fp

O.I{Ij

Fp FOR {IJ

4.0000

0.05211360

0.05211360

5.0000

0.034682~0

0.03468210

206667

0.10660515

0.10660515

3.3333

0.01251348

0.07257348

2~0000

0.11047258

0.17047258

2.5000

0.11915271

0.11915271

l.60CO

0.23152143

0.23152143

2.0000

0.17084157

0.11084751

1.1429

0~36454339

0.36454339

1.4286

0.27176150

0.21776750

1.3333

0.30315946

0.30315946

1.6661

0.22453549

0.22453549

1.0000

0.42021823

0042027823

1.2500

0.32883289

0.32883289

0.8889

0.46996619

0.46996619

1.1111

0.37668549

0.37668549

10

Oc8000

0.51380140

0.51380140

1.0000

0.42080171

0.42080111

11

0~7213

0.55221129

0.55227730

0.9091

0.46102845

12

0~6667

0.58600206

0.58600207

0.8333

0.49745186

0.49745186

13

006154

0.61559419

0.61559415

0.7692

0~53029763

0.53029767

14

0..,5714

0.64162979

Oo64162,,}80 -

0 .. 7143

0.55986124

0.5598612 8

15

005333

0.66462109

0.66461995

0.6667

0.58646125

0.586463 18

16

0.5000

0 .. 68500707

0.68501309

0.6250

0.61043230

0 .. 61041 90 3

17

004706

0.70316112

0070317893

0.5882

0.63201149

0.632041 55

18

0.4444

0.71944314 _

0.71953328

0.5556

0.65129.948

0.651581 67

19

0.4211

0.73389956

0073430325

0.5263

0.66949999

0.6693728 0

20

0.4000

0.15044759

0.74616106

0.5000

0.68565416

0.68427083

NUMBER OF MESHES IN MARCHING .OIRECTION

6.36

~~o
I

L FlIGI-IT

A t-

-'j. \

I
I

I
S_I

],"'-c

S+-I

r.

T
I

)(l

I-i!t.

!-~J

,ERO

;J.r

_'J.4." .

)(.

T
I
(

'....,

')($"

11!3

"1-

lil

.,,'

l~~

-')I. (,

"

7..
,

')(

l',

,, ~ ,

E.

)(

, ,

"

'
'

,'.

6.37 '

case considered.

Atypical region with boundary conditions '

for'pressure back flow is shown in Figure 6.12.

' direction is from E to Co

The marching

The discussion ' is restricted to the

pressure back flow casea

The matrices satisfy in zone ED, recursion equation of type
(609}e

Thosein the oblique zone DG must be determined in a

different, wyo
,' Unknowns,xk (k = l,n) are introduced at i = 2 (Figure 6.13)"

= 1 then

Frqm symmetry at i

Let

j " (j = 1, n ) bethe set of function values at i = Ie

This

set- can be expressed in xk,O and xk ,2 from finite difference

equationso

For instance, for , square meshes and n = 7

4r
-

cl

.- "'21.

+ 4 ,7

l 2 -

2 ?2. . = X, - 'X7
~ :>

(b.s"! )

::.

+ 4 '2">

The coefficierit-s on the left side of these equations form a

..:
.
tri-diagnal m~trix~ Any , zj oan with the routine of Th~mas (12)
be expressed in xko

The matrix

fE} 1

is then known while

f EJ 2

is the identify matrix o

The matrices in zoneED, which'are ,independent of "the helix
angle and independent of the channel ratio Leo/bo, are calqu- ..
lated in a separate computer programo

These matrices are square,

do not contain constants, and are antisymmetrical in ' this sense

that the

element~

satisfy the relation

6.38
The

ma~rices

program.

for zone DC are calculated in a second computer

Two previously calculated matrices are there used

as initial matrices.

Let fEls-l and fE}s be these two matrices .

= s must correspond to the index at point D of

The index i
Figure 6.12.

The function is at the boundary BC equal to the constanty.;o .

It is at D equal to the linear equation in xk given by the
coefficients ' of the last row of [E}s.

The boundary condition

is then

where " e (n,k) are elements of fEls.

of n - boundary equations.
eliminated.

This equation is the first

One_unknown' could at this point be

That elimination is, however, postponed until all

boundary equations have been determined.

The calculation of { E} s+l is similar to that of previous
matrices.

The range for j is now reduced by . l.

The matrix

tE}S+2 has that same range.

Calculation of the (n - l)th row of fE)s+2 must be based on an
equation for the irregular star of Figure 6.14.
shown with an arm

f-= 0.5.

of a different length.

'0.= . 'f.'c

11.

IfIs:: l.f c.

That star is

Other helix angles may require arms

For this star

6.39

F\<o.G,. \4 _ IR((t G u LAR- STA

Elimination o,f

fo~ POllo.JT' I'JEAe OB\-\ CS( uE

ENt>

('0 'f/o0 )c. . Yi .el~~

(~)c.= h~~ (I ..0 [\Yo +[ 4-a:' (1+ il ~e.]

while also

'.

"2.

ti

\ .

( 'X ~ ) c. ::

~ ('+0

2.

lj..tc. +

4JL.

'

The fini te difference equatiQn for the Laplacian becomes.

. . .

The matrix

(I +

{B ~

t)

g~ I 4-B - 4-

IJle. -

L -

tf. ,)+ I] ~=o

is now reduced t o a (n-l)x(n-l) matrix.

(,.ss)
Elements

of the (n_l)th row are .

t("Y\-I)'Y\-l)
'.

The {n_l)th row of

=-

1El+2

-f-+1 ).

b('Y\_I.>'Y)_I)~ ~(j; f).

S

y{elds the second boundary equation.

The irr~gular star for the last point C (Fig. 6.15) is again
different since the derivative vanish.e s along LR.

t-ro '::. ~c.. . +1.~'/..h'/..(~)

2
,
() '1.1. C.
while alsO

])'L'f)::
h'1. C
OX:L.

i:..

Lu

Ta -

2:

(IJ

Tc.

(IJ

rL

Here

6.40
t>\SCHARG'E END Wl'Tl-\ ~o=4-'o ~
=Co~Si.

~~ .

9:t

0., =0

,FL \ c:; 14 T_ _

FI<.

G.

Fo"

----t-~ l--!--------<-J----~

IS;--

IQREGULA,C2. STA.R N~Arz. ~D O'F RG61QN

VN IFO 12"",
.,;.

pe ~ S vl2 rs 4 L..O AJG,

t )~ -~ ~

I'f1-:0
11)'1

EkJl:>

,If'p,-If[ - 'Ii - f +I) Cf-:, = 0

2.(1

1) JSCI.{~(G~

./.

2 (I +

+-1-) 11.;-

$~

tl.J

't'~. ~:,- -

.",

'YR, -

2
f2.

(Ij

To:=

6.41
~'

' The'

rinite ,. d~fference

" p imilar

' equa~ions

. Fi'g ure 6. 16.

equation for the Laplacian becomes

for . smaller helix angles are shown ' in

.'

The n-,b 01:lndary equations are finally co11e c't ed

in one matrix equation

Th~ ' ~olu ti on

of this set of

eq~ations

yi e lds xk.

The , flow rata is by Simpson's rule expr~~sed as

~ 63>< J><+,;~+4(')(l+-X\,\_,)+ 2.(X3+X~~ ..)+----

'q,

+ 2 (X ....... + )(1) :~

'=

~ ~)<

[ )(.,

+ 4 x .. 4-

2.)(~ +

4 (')(">1_ ' +

. - ~ - '. - .j.. '2 '-<,, _,-+4

..
'. Equation (6 . 60) is similar ..to (6,,30 ).

t~e . ~ow vecto~

cienis: df

....

x~) ~

X'\.. ,+

'l< .... _, of

'I<.,l

~~ J

((,. C:; 0)

The flow rateco.effi-

fFJ are for this case' very small e

c' "

"

1;-

Errors i n xk are then,not propagatd in the c~lculation of

end -' effect . factor.s .. This " is not true for the cal,ulation of
the functionin the region . ;The coefficients in

fE} i

ari increase similar to that of Figure 6.6& , Small

sho~, !

e~rors

can

then cause large errors in the calculation o+' ttle function,

"

', : . ', particularly

: when
i is
large .
.
.
.
~,.

'

. ,

This ' is t rue

i n ' Figure 4&l5

'1Ii;~ ..

.'., where back substitution of x k yielq.s funct,ion values ' in the

"

tip of the region that do not' mai?c}:l the original cons't1ant

,

pressures "at the boundary

o, The function values are here
.
"

.'

correct at 'points -where th's coeffi ,c ients 'o f

1
.
[ E f i are small e

The difference between the exact and calculated end effe ct

factors is again a function of the meshsize h.
for derivatives have errors of theorder of h 2

Equations
The final

error is a combination of errors of the order of hand h 2 ,

since the region has a singular point at B (Figure 6.17).

--~

'Q

~~~~~

I\~ /"1
'
1
'1>'1 ,=0
~ h
I.
I',

"

~I

I"

I"

I
I

I
I

,
'
' !t~~
,D

,?'-, -+--?---o--". I
i

FIG.;". '7

.
I

.r-0'

El<l2.oR.. "'R.oPOf<T\o/..J~l- Tc

l"-l '?OI.,Jt: (Qg'

The deri vati ve must vanish along the 's ides of the Qhannel.
The ' function must be constant along the
requirements cannot then

b~

boundary. -;,These

satisfied at B.

issatiSfied ' at Pand Q but not at B.

The drivative

This means that the

derivative between Q and B-:dbes not vanish.

l'he finite

difference method yields an addi,tional flow ratre across the

cbannel wall betwen Q and B.

That flow rate is of the order

of h.
The calculated end' effect factors were for that reason
f .

analyzd by : the least squares method using a second o:rder

polynomial in h as a model.

The extrapo.lated

re.~.~lts
1

used to d:'e termine the coe'ff'icients. ' Cjo

(4.67).

'

were

6.43
6.9

Additiortal Examples

A number of additional problems were solved by the same technique.

For instance, a program was developed for. the solution

of Poisson' s equation in the arbitrary region of Figure 6.~ .

This program was developed to calculate pressure gradients
forviscous flow through tubes of any cross section.
Shape factors FP (6.31) were wi th this program deterIllined !or
a number of cross sections with known analytical solutions
for p

8.

Table 6.6 lists some of the results.

Table 6.6
!!UJ m(;2) Ep - Calculat"ed

Contour
Isosceles Triangle
Semicircle
,Full Circle
Square

7
9
9
6

-6
4
9
6

0.099764
0.447328
0.295039
0.413476

Ep - Exact;.
t

0.1000 (ref. 10)

o. 444 ( ref. 10)
0.2845 = 31( /28
0.4217

1) Number of unknowns.
2) Number of steps in marching directiono

. Improved results could have been obtained by extrapolation

or calculation at higher precision.

The results are withol\lt

theserefinements weil within the accuracy normally required

for engineering calculations.
- Another program was deve1:oped to simultaneously sol ve

Laplace' ~)

'and Poisso.n' s equations in the same region ,for different

boundary. c.o nditioris.
drag

fl'~

This program was developed to calculate

and pressure back flow coefficients for channels

,":

with one flat slde and a symrnetrical contour for the remainder
,

of the .:boundary.

b.44
Laplace' s eqp.ation goverl1s the drag flow. , . Here the flat
.~

side has a uniform velocity in the channel direction,

the remainder of the boundary is at a standstill.
equation governs pres's ure back flow.
at 'a standstill for that case.
tions differ

~nly

in the value

in Laplace's equation is zero.

w~ile

Poisson's

The whole boundaryis

The two differential equaof~he

co~stant.

The constant

Boundary conditions differ

along the flat side of the contour.

The matrices were for this case

augme~tedto

accommodate
,, '

,' Goefficients for xn+l and x n+ 2 The coefficients for xk ' "/;.'
(-k = 1, n) 'are in
1i pre-~iselY the same fO,r both differeri-

.;.

fE

tial equations.

The coefficien:ts for xn+l account for the

'Poisson constant; those for x n+ 2 accommodate the coefficients

for the boundary velo city for drag flow. The unknowns Xk(L)
for Laplace and xk (P) for Poisson can wi th Crout' s method (22,)
be determined simultaneously.
Some problems require solutions in the same region for different boundary canditions.

These different solutions can almost

siIllultaneously be obtained whe,n the 'boundary matrix is gener~ted'

at that boundary.

This method was used to determine the

tmperature distributionin a hollow cylinder (Figure6.18.)

6.45

I~
~

T .... o

, ,i; = I .

. "T~q

T=o

\'"

; - IN SLlL~Te:J)

.
S

,I

'

()T

"l'{ ' ~

--

--

-qT.T

'( :4

T,. I.,

I
I

!
T:.o

T-::.T"
.

-L-=:-~

~
I -

\.

I
I

-, )
~eo
'(

T:-o
.

1''''''0 ...

f\ G. 6-. \8 - $TE..oD~

$-rATC=

"'T~ R.o V<=>~ .

PI'4R,\.0, \. ~'1. IN Sv L'-\ TE C C'1I..JJ..JoEi2..

Ho LLO W

\...'~' OJI'J Due.. 'T I-oN

,',' .

The inside of the cylinder i s partly insulated

'. of the interior is kept at a uni form
~he ~xternal ' boundary

t~mperature

Th remainder
of IOOC ...

is he l d at a uniform tmperature of Of G

The influence of the length of the internal ins'ulation on

the temperature distribution is

determ~neL.

- .

A material is assunied wi th different thermal conducti vi tie.s

in the radial and i n the axial direction.

The temperature

distribution must satisfy

wher Ck is the rat i o of thermal conductivit ies.

Unknowns xk (k

= l,n)

are introduced at i

=2

close to the

outside radius. Matrices are calculated in the range i

=3

to m + 1 where m corresponds to the internal radius. Boundary

conditions are

These conditions yield n-equations in the unknowns xk and the

temperature To for any value of s.
with a mesh point.

Here z

=s

must coincide

The same matrices can be used to calcu-

late different sets of unknowns xk for any acceptable value

for s.

Figure 6.19 shows a typical temperature distribution

= 2.

for Ck

The temperature gradients d

region.

TI 0 z

are small in part of the

Calculations can in that case be significantly

reduced using a larger mesh size in th at part of the region

where gradients are expected to be smalle

This program was

designed to accommodate different mesh sizes in the z

direction.

Figure

6.2~

shows the calculation results for two

different sets of mesh sizes.

for n

= 12

The problem was first solved

with a square mesh, then for a combination of four

single and four double meshes.

The underscored temperatures

refer to the last calculation.

The minute temperature

differences show the utility of that approach, which reduces

the calculation time by a factor of approximately 3.

6.47

h
::.

-011
A

1 \1

\ 1\~ 1~ \~ \
~\ ;;\ i\
\
\ \ ~\ 1\ \ \
\ \ \ ~ \\ \ \\ .\
\ f\ 1\'~\l\ ",,\~
"
\ '"~ ~~~K
\

..."

In
1\

'""

-11
r

~--'

=>--,

cDO

0::0=:

Ln\

t'I

(\}

<t

("I

()

~~

-0

0
0

0--,
W--'

1\

0::0

=>I
~z

0:-

11

~ r-..........

~
~
~
r-~r---- i'--............
r---"-

1'---1

a..
~

,-.--

tI

0'>

.
.
<9
(j)

LL

.J

.,..:>
7

FI-G-.6.Z0-INt-lUU-JCE CF INCREASED HESH ~IZE ON T:t1PERJHUKE OISTRIBUTICN

o.

o.

0.-

o.

o.

o.

0.

o.

o.

0.-

o.

()'\

o.

,f-'(..

o.
o.

'0.73

t o.

11. 61

tOe

h.6b

0.78

o.

12.73
12. 79

:l

IL~~_8.6 7 __ ~ ? _._.~Q._.~_~q_.}~~1_~~?4_ ..~ ~1 .~.3~~.. _.-!}._~9__~ 11.67

3. 30 -l~~4~_?~~
l. Z8
-- -f--

3.31J'lt.86.

H.(lH

,: 11.3-'t
-- I

j lO.37 f
I
I

'

---

.
l
'
I
:
lJ:5.3~0 .~-{ 3~'l12.25r_~_/i~1~lS.49J
. l.6_~_?_~_ .. 17~~~,~_!_~!..~_~.~5
91
i lO. 2 2 :
1 4 0 :i
1
6 if :
1 U ',
-- 1-- 1
I
!:
rj It

H~

iJ

-1

c.

1'.... u?

'I!Y~

1!.L..!.l}.

i~~-:Q!'

i
i~O- :!?
I
;
i
;

I ., --.--...--....e
-o-----.j-:---::--v-! '.rr' -l' .. .- n.----.---::
! 5. Tl

O.

o.

O~--- -

11
.] 1.

?1. ? ('
1 _S

I 1 6. SC}
l(, 6

I i.'

i 22-. 9-9

~ 1.")

( <j 1

'3'i .

7"J

~',

. 1 r) 6 (,

1..

'j

- , (::~.i --- l f~;; <iC,n ~ ~~_~ T- --{ ~~~~~ii - i"

,

<):)

3-1. 28

2 ti. I 0

3 ,). 1 L

: '12. 3

i - - -- ---

i 33. 31
3J

79

') -

3 /t
- c---- -;

1_

0 G. . _ {,,,. '; o

() ._~J J

) () I.J

(,0.) (

)O().

1 (, J I>

.~ ' (H _i~;

l ( ./d '

:1 ') / (.

JU O.

18 67
,

i ~l>~

I !

- 35. 80 ! 36. 38 f 36. 57

r: ---:;- :- - - - - -,- - - ,
3 '.). IL '
, j b :> tJ
I
! ~

---- .., .:

---.-..

- ---

~ ~ I: {..8 ! I, {>. 3 -1 1;'( 'r 2 !'18 0 /, : 4 EI. 25

-.--- ., ~1~_~' 6i-T---- --'--.-- '--'- 4]~._-~- ,.~ --------i ~ u. 2 ~_

:. :

or

'

II

21

i ~!~. )- !- :

i: ) (JO.

I 100.

_._~

) ti 0

, '; )

. B ?

: ~?(!-'- I

I 100.

'

.!n. 'r, _,I';. 5 " :"

JO U .

~-~~--? ,~- r

_,

- ---_.--

j: 2-~~11!J;:K ~~ ~~; --if.: ~:;;: -1"0 ,~ ~ - i~ ~: ~ ~6. /] . ~:U~t _;6.

II'J

o.

12 B 2

I
!

1.1

-----...

--- - -- ;3~! ~ :.~i:~!

..9~o. _______. h/t.:~

I!!~~ -o. .~---t~lJ._~_:~.
!:' .___ ?~--t "" .,J y.o..,
.l.('. I~~~
~ O.

1 2L~:_~~_

-<'-:;--_- -. . . -- - \. -- ------- " - . -:;

(I

g~--

:25.2

-- -'- 1------ . ,- --- .- i--.. --

Ij.

'-

T~{ :lt- l~-~ !~-i ~ =_:~-~

17 n 90.
I B.

.l.l .0q . _____ _1

t}

1?_9,--~,_! 79. 37

1 /9 ._3't -

OO_~. ___ Jl.95~ ~ ____

1 00.

! 100.
Hl!).

I
SUMMARY

Screw pumps are used to pump very viscous liquids. The walls
of the helical screw pump channel are formed by two cylindrical surfaces, the barrel and screw root surfaces, and by two .
helical flight surfaces. These channels have oblique inlet
and discharge ends. This channel geometry is considerably
simplified in the conventional theories for flow of viscous
,liquids through screw pumps.
A straight prismatic channel with a rectangular cross section
is assumed in existing theories. The true geometry of the
channel ends is ignored. Identical velocity distributions
are assumed in parallel planes perpendicular to the channel
direction, which result in uniform pressure gradients in the
channel direction and in the direction perpendicular to the
flights.
Flow through rectangular channels with large aspect ratios
is, in the Simplified Theory, based on flow between parallel
plateso Theinfluence of the short sides of the rectangular
cross section on the velo city distribution can be neglected
in that theory . The resulting performance equation is useful
as a reference equation. The drag flow and pressure back
flow rates of other theories can be compared to that equation
by the introduction of corre,cton factors.
Shape factors are introduced in the rectangular channel theory
to account for the reduced velocities along the short sides of
the cross section. The true cross section is, however, not
rectangular. It departs more from the rectangular cross
section the de eper the channel.
The actual channel cross section is used in the analysis of
Chapter 28 Here performance equations for isothermal flow of
a Newtonian liquid are compared with those of the simplified
theory for screw pumps with large LID ratios, where the channel
end geometry can be neglected. Congruent velo city distributions are again assumed. An analytical solution is given
for the velocity distributions in helical channels with large
aspect ratios. Correction factors FDC for drag flow and FPC
for pressure back flow are derived as functions of the helix
angle and the relative depth.

11

The differential equations cannot be solved for flow through

deep channels. The correct ion factors then do not account
for the reduced velocities near the flights. These reductions must,for the helical channel, be quite similar to those
of the rectangular channel theory. Shape factors for the
helical channel are approximately equal to those for rectangular channels with a width equal to the average width of
the helical channel. The effect of curvature and flights is
shown as combined factors FD2 (Fp2 ), which are the products
of these shape factors and FDC (FpC ).
The combined factor for drag flow, FD2 , is always smaller
than l.O. Drag flow is smaller than predicted by the simplified theory, but it is also smaller than that of the
r~ctangular channel theory.
This factor decreases for all
,helix angles with increased depth. That decrease ~ is smallest
at ~~ -= 20 0 Smaller or larger helix angles 'show "'a more rapid
decrease with depth. The similar factor Fp2 is about 1.0 for
'fo = 20 0 and is almost independent of the channel depth for
tht angle. It decreases for smaller or larger helix angles
" Wi,~h increased depth.
The decrease is more rapid the more the
',<, h-el.ix angle differs from 20 0 The pressure back flow rate is
': ,larger than predicted by the rectangular channel theory. The
. e'nergy dissipated" in a helical channel is also less than the
one derived in the simplified theory.
..

The conventional leakage flow analysis is for isoth~rmal "flow

extended in Chapter 3. The leakage flow consists of one part
proportional to the screw speed, another part proportional to
the dis charge pressure. The magnitude of the leakage flow
rate is influenced by the screw geometry, the mode of operation
of the screw pump, and the ratio of the viscosities in channel
and flight clearances. Non-Newtonian viscosities can have a
: large viscosity ratio, which increases the leakage flow rate
significantly.
The leakage flow rate causes formation of a thin wall layer
, on the inside of the barrel. The' thickness of that layer is
influenced by the same parameters ' that influence leakage flow.
That layer plays an important role in heat transfer. Heat
transfer coefficients must then be based on the thickness of
this layer and not on the flight clearance, as is done in the
conventional theory.

111 -

The inlet and dis charge pressures of screw pumps are substantially uniform. Only for one operating condition can the
uniform pressure gradients of the conventional theories
satisfy a uniform pressure at the oblique inlet and dis charge
ends of the channel. Isothermal flow of a Newt.onian ~iquid
through channels with oblique ends and large aspect ratios is
discussed in Chapter 4 for uniform inlet and dis charge press.ures.
The average channel velo city components are there' expressed '
as the derivatives of a velocity potential. That potential
can be regarded as the sum of a potential for pressure ba,ck.
flow and one for drag flow. The drag flow potential can again
be separated into one caused by the barrel velocity component
in the channel direction, and one caused by its transverse
component. The boundary conditions for each case are simple..
The potential for each case must satisfy Laplace 's equat.ion.
These equations are sol ved by numerical techniques. The end
e.ffectfactors for drag flow, FDE , and that for pressure' back
'flow , FpE , are determined for a series of helix angles and a
series of L/D ratios. A simple relation exists between FDE
and FpE , which shows that the factor for drag flow cannot be
important for the usual screw designs.
The ends cause a pressure distribution that differs from that
of the simplified theory only in a small region near the
oblique ends. The transverse pressare gradients vanish for
, the pressure back flow case in parts of the channel removed
from the oblique ends. Regions without significant transverse.
gradients exist when the channel length exceeds. a critical .
length . Equations are given for FpE for channls with a =.-;.\.
greater than critical length. The end correction factor fo11"
pressure back flow is significant for the usual helix angles '
and LID ratios.
Experiments, carried out to measure the end effect in chann.e ls
with oblique ends, are described in Chapter 5. It was not
possible to measure end effects in rotat.ing equipment. A
stationary model was built to simulate a s crew pump chann.e l
with a helix angle of 18.5 0 and a channeI rat,io LCD/bO :;.:. 5.The length of that channel is about equal to tl+.e criti.cal
length.
..
The difference between the theoretical and experimental end
effect factor was o.f the order of 2%. Streamlines were made
visible by injection of colored corn syrup streams in a clear
main stream. Observed streamlines diff~red slightly from
calculated streamlines. Seventeen stand pipes in the oblique
end were used to measure the pressure dis tri but ion, which
agreed weIl with the calculated pressure distribu~ion.

IV
A new numerical technique was developed for the sol ut ion of
second order partial differential equations in families of
regions for the same boundary conditions. Some of the
elements of that method ~lere des cri bed by Courant before
electronic computers were used, but it appears that this
methbd was never fully developed. The technique, which does
not require iteration, is described in Chapter 6. It does
not require evaluation of the function in the region when a
flow rate or other integral function is the only desired
result.
A limited number of unknown function values are introduced
in this method at mesh points near one boundary. All other
function values can be expressed as linear equations in these
unknowns. The function values are equated to known boundary
values at the opposite boundary. The number .of boundary
equations is equal to the number of unknowns. Similar
equations can be formulated when the derivatives at the
opposite boundary are known. The unknowns are then solved
from a set of linear boundary equations.
Flow rates and other integral functions can, by Simpson's
rule, also be expressed as linear equations in the unknowns.
These integrals can be evaluated once the unknowns have been
solved. Evaluation of the function in the region is, in that
case, not required.
'
Theset of boundary equations can be expressed as a matrix
equation. Matrices tend to become ill-conditioned when the
number of meshes between opposite boundaries becomes large.
Thus solution is only possible when calculations are performed
with the required precision.
The influence of precision on the stability of the calculation
is demonstrated for viscous flow through a tube with a
rectangular cross section. The numger of meshes, for which
the calculation becomes unstable, increases with the precision.
That number is, for this example, independent of the number of
unknowns
Linear equations for integral functions and for function values
have alternating positive and negative coefficients that increase rapidly when the number of meshes between boundaries
increases. Errors in the unknowns are then propagated in the
evaluation of these functions. This does not apply to the
calculation of end effect factors. The coefficients in the
flow rate equation are there small since that equation is based
on the derivatives at the boundary where the unknowns are
introduced. The magnification of errors in the unknowns does
not occur here.

v
The calculation of shape factors for flow through tubes with
rectangular cross sections and that of end effect factors is
shown to illustrate the methode The method is much shorter
than existing iteration techniques when integral functions
must be evaluated in families of regions.

VI
S A MEN V A T TIN G
Schroefpompen worden gebruikt voor het verpompen van hog
visceuze vloeistoffen. Het schroefpomp kanaal is de ruimte
tussen de pompcylinder, de schroefkerncylind~r en de twee
radiale schroefvormige oppervlakken van de schroefdraad.
Dit kanaal eindigt aan de inlaat en uitlaatzijde met een
scherpe punt.
De exacte kanaalvorm wordt in bestaande theorien niet
gebruikt. Het kanaal wordt daar altijd vervangen "door een
recht prismatisch kanaal met een rechthoekige doorsnede.
Men neemt dan aan, dat de snelheidsverdelingen in alle
doorsneder.. loodrecht op de kanaalwanden congruent .. zijn.
Deze veronderstelling geeft uniforme drukgradienten, ' zowel
in de kanaalrichting als loodrecht op de schroefdraad.
De schuine einden hebben in die beschouwing geen invloed
omdat de druk aan de kanaaleinden nu vastgelegd is door de
drukgradienten.
Kanaaldoorsneden met grote aspect verhoudingen worden in
de Vereenvoudigde Theorie behandeld. ,Snelheidsverdelingen
zijn daar, onder verwaarlozing van de invloed van de korte
zijden van de rechthoekige doorsnede, gebaseerd op verge lijkingen voor visceuze stroming tussen evenwijdige platen.
De hieruit volgende vergelijking voor de pomp karakteristiek
kan als een standaard vergelijking gebruikt worden. De schuif
stroming en de drukstroming als gevolg van de drukgradient,
die uit andere theorien volgen, kunnen door invoering van
correctie factoren uitgedrukt worden in de overeenkomstige
grootheden van de Vereenvoudigde Theorie.
Vormfactoren worden in de rechthoekige prismatische kanalen
met eindige aspect verhoudingen gebruikt om de invloed van
de korte zijden op de opbrengst weer te geven 9: De doorsnede
van het schroefvormige kanaal is echter niet rechthoekige
Het verschil met een rechthoekige doorsnede wordt groter
naarmate het kanaal dieper is.
De invloed van de juiste kanaal vorm op de schroefpomp
karakteristiek wordt in dit proefschrift geanalyseerd voor
isotherme stroming van Newtonse vloeistoffen. De juiste
kanaaldoorsnede wordt in hoofdstuk 2 gebruikt. Karakteristieken worden daar berekend op grond van de veronderstellingen,
dat de pomp een grote LID .verhouding heeft en dat , de snelheids verdelingen' in doorsneden ~oodrecht op de schroefas congruent
zijn. De invloed van de schuine kanaaleinden is hier
verwaarloosd. Een analytische oplossing kan slechts gegeven

,-

. "

VII
worden voor kanalen met grote aspect verhoudingen. Correctie
factoren FDC voor de schuifstroming en FpC voor de drukstroming
worden berekend. Deze factoren geven echter niet de invloed
van de radiale kanaalwanden op de stroming.
De vormfactoren voor een kanaal met een rechthoekige doorsnede
zijn ongeveer gelijk aan de vormfactoren voor het schroefvormig kanaal. De gecombineerde invloed van de radiale wanden
en de ronding van het kanaal wordt dan benaderd door het
!
gebruik van 'gecombineerde correctie factoren FD2 (F p2 ),
die gelijk zijn aan de producten van FDC (FpC) en vormfactoren.
De schuifstroming is kleiner dan die volgens de Vereenvoudigde
Theorie, maar ook kleiner dan die voor prismatische rechthoekige kanalen met eindige aspect verhoudingen. De f~ctor FD2
wordt voor alle schroefhoeken kleiner voor diepere kanalen.
De invloed van de kanaaldiepte is het kleinst voor 'po = 20 0
De factor Fp2 is voor ~o= 20 0 ongeveer 1.0 en wordt dan niet
sterk beinvloed door de kanaaldiepte. Deze factor wordt voor
andere hoeken kleiner voor grotere kanaaldiepte. De invloed
van de kanaaldiepte is groter naarmate de schroefhoek meer
verschilt van 20.
De bestaande analyse van de isotherme lekstroom is. in hoofdstuk 3 uitgebreid. Deze lekstroom kan verdeeld worden in
een gedeelte dat evenredig is met de schroefsnelheid, en een
gedeelte dat evenredig is met de druk. Beide gedeelten zijn
functies van de afmetingen van de schroef, de bedrijfsvoorwaarden en de verhouding van de viscositeit in het kanaal tot
die in de schroefspeli ng . De lekka ge wordt aanzienlijk
verhoogd wanneer de niet- Newto ns e vis cos iteit in de speling
kleiner is dan die in het J chr oefka naal.
De lekstroom vormt een dunne vloeistoflaag aan de binnenkant '
van de pompcylinder. De dikte van deze laag wordt beinvloed
door dezelfde factoren. Deze laag speelt een voorname rol in
warmte overdracht. Het is dan onjuist de coefficienten van
warmte overdracht te baseren op de spelinf: , zoals dit gedaan
wordt in bestaande theorien. De coeffic ~_ enten moeten
gebaseerd worden op de dikte van deze laag.
De invloed van de schuine inlaat- en uitlaateinden van het
kanaal op de pompkarakteristiek wordt in hoofdstuk 4 behandeld
. De dr~kken aan die einden moeten vrijwel constant zijn. De
bestaande theorieen, die gebaseerd zijn op uniforme drukgradienten, geven alleen in een uitzonderlijk geval uniforme
einddrukken . De nieuwe theorie behandelt isotherme stroming
door rechthoekige kanaaIen met grote aspect-verhoudingen,
waarbij het kanaal eindigt in schuine einden.

'.

VIII
De gemiddelde snelheidscomponenten in het kanaal kunnen
worden uitgedrukt als afgeleiden van een snelheidspotentiaal.
De totale potentiaal kan worden beschouwd als de superpositie
van een schuifstroming potentiaal voor de cylinder snelheidscomponent in de kanaalrichting, en voor de cylinder
snelheidscomponent ' loodrecht op de schroefdraad, en en voor
de drukstroom. Elke elementaire potentiaal moet voldoen aan
de v~rgelijking van Laplace voor eenvoudige randvoorwaarden.
De

~orrectiefactoren

FDE voor de schuifstroming en FpE voor

"de drukstroming worden berekend met een numerieke methodee
Deze factoren zijn functies van de schroefhoek en de relatieve
lengte van het kanaal. Er bestaat een eenvoudige relatie
tu~sen F
DE , FpF. en de schroefhoek, waaruit men af kan leiden,
dat .de scnuifs~roming niet erg beinvloed kan worden door de
schuine kanaaleinden. De snelheids verdeling in een kanaal
met sch~ine einden is voor gelijke opbrengst alleen verschillend van die volgens de Vereenvoudigde Theorie in gebieden
dichtbij de inlaat en uitlaat van het kanaal. Een formule
kan afgeleid worden voor eindeffect-factoren voor kanalen
met een lengte groter dan ,een critische lengte. De factor
.
voor de drukstroom is van betekenis in schroefpomp berekeningen.
Experime~teri voor de verificatie van het eindeffect zijn
beschreven in hoofdstuk 5. Het was niet mogelijk eindeffecten te meten aan roterende machines. Een schroefpomp
kanaal met een lengteverhouding LCO/bO
5 en een schroefhoek
0
van 18.5 werd gesimuleerd in een modelkanaal. Deze
kanaallengte is iets groter dan de critische lengte. Het
verschil tussen de theoretische en gemeten eindeffect factor
bedroeg 2%. Stroomlijnen konden bestudeerd worden door
injectie van gekleurde stroov in een transparante hoofdstroom.
De gemeten drukverdeling en gemeten stroomlijnen kwamen goed
overeen met de berekende waarden.
"

Een nieuwe methode werd ontwikkeld voor het oplossen van

tweede orde lineaire partiele differentiaalvergelijkingen in
families van gebieden voor dezelfde randvoorwaarden. Deze
methode is beschreven in hoofdstuk 6. Sommige elementen
werden reeds voor de invoering van electronische rekenmachines. "
door Courant beschreven in een college dictaa~maar de methode
werd blijkbaar nooit ontwikkeld tot een bruikbare method~.
De methode vereist geen iteratie. De functie hoeft niet
overal berekend te worden wanneer alleen de integraal van de
functie het verlangde resultaat is.
Een beperkt aantal onbekenden, gelijk aan de functiewaarden,
wordt in deze methode ingevoerd in knooppunten nabij een van
de randen van het gebied. De functie kan dan in elk ander
knooppunt uitgedrukt worden als eeen lineaire vergelijking

IX
in -deze onbekenden. Dergelijke vergelijkingen worden dan aan
de tegenovergestelde rand van het gebied gelijk gesteld aan
de functiewaarden aan die rand. De oplossing van eeri stel
lineaire vergelijkingen geeft dan de waarden van de onbekenden.
De integraal van de functie kan met de regel van Simpson ook
uitgedrukt - worden als een lineaire vergelijking in de
onbekenden . De numerieke waarde van de integraal kan dan
bepaald worden zonder dat de verdeling van de functie bekend
is.
De lineaire randvoorwaarde-vergelijkingen vormen een matrixvergelijking. Dez e matrices worden slechter geconditioneerd
naarmate het aantal mazen tussen de twee randen groter wordt.
Problemen kunnen dan alleen opgelost worden wanneer
berekeningen uitgevoerd worden met de juiste precisie.
De invloed van de precisie op de stabiliteit wordt gedemonstreerd voor een visceuze stroming door en buis met een
rechthoekige doorsnede. Het toelaatbare aantal mazen tussen
de randen van het gebied wordt groter naarmate de gebruikte
precisie groter is. Dit aantal is onafhankelijk van het
aantal ingevoerde onbekenden.
De berekeningen van vormfactoren voor rechthoekige doorsneden
en eindeffect-factoren voor schroefpomp kanalen worden
uitvoerig besproken. De methode is veel korter dan de
bestaande iteratieve methodes, vooral wanneer een integraal
moet worden berekend in een familie van gebieden.

x
Nomenclature
A
AO

B
br

bO
b

Cl' C2
c
)'c

c'

:=:c

G1

cp
C~
D

Dn
e
El, ECI
E .
F

Ep, EpC
ET, ETC
E

F (ot)

FD; Fp
FDC, ' FpC
FIJI' Fpl
FD2, Fp2

Fp3
FDE , FpE
F (r)

f, -fT
G (r)

G (ct.
h

Constant in eq. (2.15)

Constant in eq. (1.11)
Ooefficient matrix (2~16)
Channel width at radius r
Channel width at barrel diameter
Channel width at average diameter
Constants of integration
Radial clearance
Leakage wail layer thickness
Screw root layer thickness
Specific heat
Coefficient defined in (4.66)
Barrel diameter
Determinant of reduced matrix
Axial flight land width
Energy dissipated in channel per unit time
Energy dissipated in flight clearance per unit
time
Energy at pump dis charge per unit time
T otal energy per unit time
Matrix defined in (6.13)
Function of ()(, (2.32)
Shape factors for rectangular cross section
Curvature factors
Shape factors for helical channel
Combined correct ion factors
Back flow correct ion factor based on average
diameter
End effect correction factors
Factor defined in (2.l2e)
Flow factors, dimensionless
Function defined in (2.l2d)
Function of eX. (2 .40a)
Channel depth

XI
Energy coefficients (2.69, 2.70)

HV' HVT
h
I

K, KO
K (eX- )
kC, k~ ,k~o
kO
L

LCO
L*

L .......
',",'

Ll' L2
n
N
n,m
p, P
Po
PS' PST

Pressure difference across flights

Pressure generated in one turn

b PF
6

PT
Q, ~
QDT, QDO

Total flow rate

Drag flow rate of Simplified Theory
Pressure back flow rate of Simplified Theory
Drag flow rate
Pressure back flow rate

QPT, QpO
QD, QDC
Qp, QpC

Drag flow rate in channel with oblique ends

QDE
QPE

Pressure back flow rate in channel with oblique

ends
Drag flow rate due to Vo
Flow rate per unit length
Cylindrical coordinate
Radius at barrel
Radius at screw root surface

QDl
q
r

RO
Rl
S (<Pa) 0(,

S -

Mesh size in finite difference calculation

Identity matrix
Energy coefficients (1.58, 1.65)
Funct.ion of ~ (2. 40b )
Leakage fractions (3.32, 3.33)
Leakage coefficient
Axial length of screw pump
Developed channel length at barrel diameter
Critical channel length (4.60)
Length defined in Fig. 4.10
Defined in Fig. 4.10
Number of parallel flights
Screw speed
Number of meshes
Pressure
Pressure difference generated by screw pump
Shut-off pres9ure

Function defined in (2.43a)

Dimensionless stream function

XII
T ((POJ~)

ZD, Zp

Function defined in (2.43b)

Lead of s crew
Temperature
Barrel velocity component, x~direction
Velo city components
Average channel velocities
Barrel velocity component y- or Q direction
Volume in clearance
Dimensionless coordinate
Cartesian coordinates
Cartesian or cylindrical coordinate
Dimensionlessvelocity potential

ot

Ratio Rl/RO (2.22)

Uo

u, v, w
, v-

Vo
Vc

x,

x, y

Defined in .(3.8)
A PF/APT (2.60)

Pressure ratio (2.61)

r\~i'

0~
~~

Et=

'1
Y)IVI

-e
/\c:

foF

9
0-

ef;

4'

~o

~'f

.Q

Back flow coefficient in channel direction

Transverse back flow coefficient
Shear rate in clearance
Flight energy fraction (1.64)
e/t (3.15)
Effiiency
Efficiency in prime mover
Cylindrical coordinate
Viscosity
Viscosity of liquid in channel
Viscosity of liquid in flight clearances
Density
Defined in (3.16)
Dissipation function
Velocity potential
Hel'ix angle at barrel diameter
Helix angle at radius r
Reduced arm coefficient
Stream function

XIII
References
1.

J. C. Car1ey, R. S. Ma110uk, J. M. McKelvey, "Simp1ified

Flow Theory for Screw Extruders", Ind. Eng. Chem.,
Vol. 45, 1953, p. 974.

2.

W. T. Pigot t, "Pres s ure Deve10pe d by Vis cous Mat erials in

Screw Extrusion Machines", Trans. ASME, Vol. 73, 1951,
p. 947.

3.

J. C. Carley, R. A. Strub, "Basic Concepts of Extrusion",

Ind. Eng. Chem., Vol. 45, 1953, p. 470.

4.

E. C. Bernhardt (Editor), "Processing of Thermoplastic

Materials ", Chapter on Extrusion by P. H. Squires,
Reinho1d Publishing Co., New York, N.Y., 1959.

5.

M. L. Booy, "Inf1uence of Channel Curvature on Flow,

Pressure Distribution and Power Requirements of Screw
Pumps and .Melt Extruders", Trans. ASME, Series B,
Feb. 1964, and Trans. SPE, July 1963.

6.

J. M. McKelvey, "Ana1ysis of Adiabatic Plastics Extrusion",

Ind. Eng. Chem.,Vo1. 46, 1954, p. 660.

7.

W. D. Mohr, R. S. Mallouk, "Flow, Power Requirements,

andPressure Distribution of F1uid in a Screw Extruder U ,
Ind. Eng. Chem., Vol. 51, 1959, p. 765.

$.

Sir Horace Lamb, "Hydrodynamics" , 6th ed., Dover

Publications, New . York, N.Y., 1945, p. 5$2.

9.

K. S. MilIer, "Partia1 Differentia1 Equations in Engi~

neering Prob1ems", Prentice Hall, N.Y., 1953, p. 101.

10. S. Timoshenko and J. N. Goodier, "Theory of E1asticity",

2nd ed., McGraw-Hi11, New York, N.Y., 1951.
11. !te V. Southwe11, "Relaxation Methods in Theoretica1
Physics", Oxford, . Clarendon Press, 1946.
12. D. W. Peaceman and H. H. Rachford, Jr., "The Numerical
Solution of Parabo1ic and E11iptic Differential Equations",
J. Soc. IndG App1. Math., 3, 28-41, 1955.
13. E. S. de Haven, "Extruder Design for a Pseudoplastic
Fluid" , Ind. Eng. Chem., Vol. 51, 1959, p. $13.
14. R. E. Co1we11 and K. R. Nichol1s, "The Screw Extruder",
Ind. Eng. Chem., Vol. 51, 1959, p~ . $41.

XIV

15. Yoshiro Mori, "On the Screw Extruding Process for Forming
Plastic MateriaIs", Chem. Eng. (Japan), Vol. 18, 1954,
p. 841.
16. T. Yoshida, et. al., UStudies on the Extrusion of Molten
Plastics (2nd Report)", Chem. Eng. (Japan), Vol. 21,
1957, p. 336-374.
17. P. H. Squires, "Screw-Extruder Pumping Efficiency",
SPE Journa1, 1958, p. 24-30.
18. J. M. McKe1vey, "Polymer Processing U, John Wi1ey, New York,
N.Y., 1962.
'"
19. Shih-I-Pai, "Viscous Flow Theory", D. van Nostrand,
Princeton, 1956.
.
20. C. H. Jepson, "Future Extrusion Studies", Ind.Eng. Chem.,
Vol. 45, 1953, p. 992-993.
21. H. Hahnemann and L. Ehret, "Der Druckver1ust der
1aminaren StroffiUng in der Anlaufstrecke von geraden,
ebenen Spa1ten", Jahrbuch der Deutschen Luftfahrtforschung, 1941, p. 21.
22. P. D. Crout, trA Short Method of Eva1uating Determinants
and Solving Systems of Linear Equations", Trans. AIEE,
Vol. 60, 1941, p. 1235.
23. J., P. Hartnett, J. C. Y. Koh, S. T. McTomas, "A Comparison of Predicted and Measured Friction Factors for
Turbulent Flow through Rectangular Ducts", Trans. ASME,
J. of Ht. Trans., 1962.
.
24. R. Courant, "Advanced Methods in App1ied Mathematics",
p. 79-88 (mimeographed 1ecture notes, N.Y. University,
1941)
25. K. S. Kunz, "Numerical Ana1ysis", McGraw HilI, N.Y.,
1957, p. 292.

..

------------------~-------------- -

STELLINGEN

I.

Een schroefpomp kan beschouwd worden als een speciaal soort

viscometer. De effectieve schroefpomp viscositeit kan dan berekend
worden uit proefresultaten met behulp van de schroefpomp vergelijking. De viscositeit kan ook in een capillair gemeten worden voor
een schuifsnelheid gelijk aan de specifieke schroefpomp schuifsnelheid
17 ON / h (0 = diameter, N = snelheid, h = schroefgang diepte). Aangetoond kan worden, dat de effectieve viscositeit voor niet Newtonse
polymeren kleiner is dan de capillaire viscositeit. Dit verschil
wordt groter naarmate het polymeer meer afwijkt van een Newtonse
vloeistof.

11.

De snelheids verdeling van de schuifstroming in schroefpomp

kanalen met grote aspect verhoudingen is voor niet Newtonse
vloeistoffen niet lineair. Deze schuifstroming is kleiner dan
die voor Newtonse vloeistoffen. De schuifstroming wordt klei~er
naarmate het polymeer meer afwijkt van een Newtonse vloeistof.

111.

De door Ja cobi afgeleide formules voor stroming van niet

Newtonse vloeistoffen door schroefpomp kanalen zijn onbetrouwbaar omdat het door hem gebruikte superpositie principe niet
geldt voor niet Newtonse vloeistoffen en omdat de schuifsnelheid
in een richting loodrecht op de schroefdraad ten onrechte
verwaarloosd is in de berekening van de viscositeit.
H. R. Jacobi, "Grondlagen der Extroder Technik",
Carl Hanser Verlag, Muenchen, 1960

IV.

Aangetoond kan worden, dat men de invloed van de afrondingsstraai in de hoeken van schroefpomp kanalen niet kan verwaarlozen
wanneer de kanaal doorsnede een kleine aspect verhouding heeft.

V.
De door Squires afgeleide vergelijking voor de stroming van
visceuze vloeistoffen in gedeeltelijk gevulde schroefpomp kanalen
is niet juist voor kanalen met kleine aspect verhoudingen.
P. H. Squires, Chapter 4, "Processing of Thennoplastic
Materials", E. C. Bemhardt, ed. Reinhold, New York,
N. Y., 1959

VI.
De afmetingen van lange overhangende assen, zoals gebruikt In
roerwerken en pompen, worden meestal gebaseerd op een veilige
marge tussen het critische toerental en het bedrijfstoerental.
Het over te brengen draaimoment geeft in die gevallen schuifspanningen, die een fractie zijn van de toelaatbare spanningen.
Het critisch toerental kan in zulke gevallen verhoogd worden door
het aseind te verjongen of hol uit te voeren.

VII.
Het is gebruikelijk het mondstuk van een matrijs voor de
extrusie van dunne platen te voeden vanuit een lang ingebouwd
kanaal, dat evenwijdig loopt met de opening van het mondstuk.
Men kan aantonen dat de gelijkmatigheid van de extrusie voor
hetzelfde drukverlies beter wordt wanneer dat kanaal vervangen wordt
door twee evenwijdige kanalen met een tussen geschakelde hydraulische
weerstand.

VIII.
Sommige digitale electronische rekenmachines zijn zo gearrangeerd,
dat de berekening toch voortgezet wordt wanneer de machine gevraagd
wordt een vierkantswortel van een negatief getal te berekenen. Het is
voor gebruikers van dergelijke machines onaanvaardbaar dat de machine
niet onmiddelijk stopt en een boodschap geeft met de reden voor het
beeindigen van de berekening.

IX.
Het is meer economisch problemen, die in aanmerking komen
voor oplossing op digitale electronische rekenmachines, te programmeren
en op te lossen op de snelste machine met het grootste geheugen.
Vooral het programmeren is voor machines met een grote geheugencapaciteit veel eenvoudiger.

X.
Het door Muijderman voor het bepalen van de invloed van de
schuine einden van de groeven van axiale spiraallagers op de
drukverdeling gebruikte analogon geeft een onjuiste oplossing.
Toepassing van dit analogon tot de inlaatzijde van de groef geeft
een positieve druk correctie, terwijl toepassing tot de uitlaatzijde
voor dezelfde lineaire drukverdeling een negatieve druk correctie
geeft. De drukverdeling is niet continue, wanneer terzelfder tijd
aparte analogons worden gebruikt, respectievelijk voor het inlaaten voor het uitlaateinde van een groef. Deze discontinue drukverdeling
kan niet de juiste oplossing zijn.
E. A. Muijderman, "Spiral Groove Bearings",
Proefschrift, Delft 1964

XI.
Commercie Ie gegevens over de viscositeit van polymeren zijn
altijd gebaseerd op gemiddelde of schijnbare schuifsnelheden.
De effectieve viscositeit hangt voor stroming door machines en
': pparaten af van het stromingsbeeld. Deze effectieve viscositeit kan
dan alleen berekend worden . wanneer de bepalings methode van de
viscositeit volledig bekend is. Commerciele gegevens zijn in zulke
gevallen alleen van waarde wanneer volledige gegevens over die
methode en de berekening van die viscositeit bekend zijn.

XII.
In de USA . is de vooropleiding van aankomende studenten, door
de gedecentraliseerde , organisatie van het middelbaar onderwij s, . zo
weinig uniform dat vrijwel alle universiteiten ' het eerste jaar
besteden aan een herhaling van de stof, die in de High Schools
had moeten worden behandeld. Het invoeren van een uniforme
standaard en een uniforme onderwij s inspectie zou de achterstand
met andere opleidingen aanzienlijk kunnen bekorten.

XIII.
De schroefpomp theorie wordt in de literatuur dikwijls verward
met de extruder theorie. De schroefpomp theorie is van weinig waarde
voor het ontwerp en de bedrijfs analyse van extruders, omdat de
karakteristiek van een ' extruder vrijwel uitsluitend bepaald wordt door
de druk die opgewekt wordt in die gedeelten van de extruder, waar
het polymeer gesmolten wordt. De theorie voor het smelten van
polymeren in extruders is nog niet ontwikkeld.

XIV.
Het is voor landen met een grote export van academisch personeel van belang te komen tot een internationale gelijkstelling
van diplomas en titles.