You are on page 1of 4

A Model for Plate Clash in

Disk Refiners
Frdric Alfonsi and D.J. Jeffrey
Department of Applied Mathematics,
The University of Western Ontario,
London, Ontario, Canada
Abstract
We present an extension of the Miles--May
model of pulp flo in a ood-chip dis! refiner"
The extended model is aimed specifically at
understandin# plate clash in dis! refiners,
hich occurs if the pulp pad $eteen the refinin#
plates can no lon#er support the axial thrust on
the plates and the control mechanism allos
them to touch" The model considers separately
the pulp $eteen the $ars and the pulp in the
#rooves of the refiner" As a result of stochastic
exchan#es of pulp $eteen the $ar and #roove
re#ions, they contain different concentrations of
pulp" The pulp that is compressed $eteen the
$ars determines the axial thrust on the plates" %t is
shon that the thrust can reach a maximum value
as the #ap decreases" &ast a critical #ap, the
thrust decreases ith the #ap" This decrease can
explain plate clash"
Introduction
The aspect of refiner operation that is to $e
studied here is plate clash, also called
pad collapse" The phenomenon is $asically
related to the control of a refiner, and in the early
days of control as a challen#in# pro$lem" As a
pro$lem in refiner control, pad collapse is no
lon#er a dan#er, $ecause of modern adaptive
control strate#ies, hich monitor the effects of
plate-#ap chan#es on specific ener#y" 'oever, a
theoretical understandin# is still orth pursuin#"
The standard set-point for refiner operation is
specific ener#y, and this can $e varied $y
chan#in# the plate #ap( other operatin#
parameters, such as feed rate, affect specific
ener#y, $ut here e concentrate on plate #ap" A
reduction in the #ap ill usually raise the
specific ener#y of the refinin# action" Thus, if a
refiner is $elo its set-point, the control strate#y
is to reduce the #ap until the set-point is
re#ained" 'oever, there exist operatin#
conditions in hich a decrease in the plate #ap
decreases the specific ener#y" Therefore there is
a dan#er that a control strate#y $ased only on
this simple idea, i"e", one that alays assumes
that smaller #aps mean hi#her motor load, mi#ht
$e applied to the unusual case and continually try
to raise the specific ener#y $y drivin# the plates
closer to#ether" %n such an event, the plates
eventually clash a#ainst each other" Moreover,
since the situation evolves ith increasin#
speed, it appears to $e a sudden collapse of the
pulp pad in the refiner" )ecause of this, an
important early topic of research in refiner
control as the rapid identification of the re#ime
the refiner as operatin# in, so as to avoid this
runaay situation"
The main result of this or! is an explanation of
the shape of the #raph relatin# motor load, or
specific ener#y, ith plate #ap, assumin# that the
other parameters, such as feed rate, are constant"
The #raph is shon $elo"
There are to reasons for loo!in# for an
explanationof this #raph, even thou#h at least
one operational cure for plate clash is availa$le"
%n the first place, a $etter cure mi#ht exist, and
secondly, the pro$lem is a ne challen#e for the
theory of refiners"
The Miles--May model for pulp flo inside a
refiner has $een very successful at descri$in# the
dynamics of a dis! refiner *+," %t contains,
hoever, simplifyin# assumptions
that prevent it from addressin# all aspects of
refiner operation" There is nothin# ron# ith
this, of course, $ecause every successful
en#ineerin# theory concentrates on some aspects
rather than others" -pecifically, Miles and May
modelled the $ar and #roove pattern on refiner
plates $y includin# #enerali.ed friction terms in
their e/uations" Thus they introduced separate
tan#ential and radial coefficients of friction to
reflect the fact that it is easier for the pulp to
move alon# the #rooves rather than across the
$ars" To ta!e the $ar and #roove features into
account more fully, 0an, 1effrey and Ouellet *2,
have derived a model to complement the or!
done $y Miles and May" %t is $ased on the
folloin# assumptions 3
Discontinuous medium 3 the pulp
consists of individual flocs of ood
pulp travelin# throu#h the refiner"
The refiner is divided into 4 re#ions 3
#ap, stator and rotor re#ions" The
individual flocs move from one re#ion
to the other"
The floc velocity u depends on the
re#ion" %f v is the 5Miles-May6 velocity,
then u7v in the rotor, u7+82v in the #ap,
and u79 in the stator"
:ach floc is individually folloed throu#hout the
refinin# .one" To model is advanced $y dividin#
the time into e/ual time steps" At each time step,
the state of the refiner is defined $y all the flocs;
radial and re#ional <#ap, rotor or stator=
positions" The individual flocs ill move from
one re#ion to another accordin# to a pro$a$ilistic
la" The aim of this paper is to continue and
extend this 0an-1effrey-Ouellet model to
investi#ate plate clash"
Modified Stochastic Model
We divide the interior of the refiner into 4
re#ions" The #rooves on the rotor, the #rooves on
the stator and the #ap $eteen the $ars <the
refinin# re#ion=" This is illustrated $elo"
The pulp flocs move $eteen the re#ions
stochastically at each time step" %n *2,, the
pro$a$ility that a floc moved from one re#ion to
another as fixed" 'ere the pro$a$ility depends
upon the density of pulp in each re#ion, and on
the cross-sections of the re#ions" The full details
are #iven in *4,( here e #ive a summary" The
pro$a$ility of a floc movin# from a source
re#ion to a tar#et re#ion is assumed to $e a
product of to terms" The ratio of the cross-
sectional areas of the tar#et and source re#ions
and the difference $eteen the actual local pulp
density and the maximum possi$le density" %t
should $e noted that this local density varies ith
position in the refiner, $ut it can $e calculated
usin# the data of the simulation"
Compression of the Pulp Pad
%n order to calculate the relation $eteen the si.e
of the plate #ap and the specific ener#y, e
assume that the axial thrust is supported $y the
flocs in the #ap re#ion, and in the first instance
i#nore the contri$ution from steam pressure" %n
the short time that a collapse occurs, it seems
unli!ely that steam pressure could chan#e
si#nificantly" We ill fix the plate #ap, and our
#oal ill $e to compute the specific ener#y and
motor load in terms of that #iven value"
The flocs $eteen the $ars are compressed" We
use the Compression :/uation from 'an *4, to
calculate the load each floc can support" The
e/uation is an empirical relationship $eteen the
pulp pad density > and the applied stress ?" %t is
ritten as3
> 7M ?@A,
here M and A are constants descri$in# the pulp
properties" 0or our simulation, e used M74 and
A79"4BC, correspondin# to $lac! spruce"
Floc Behaviour in the Gap
We assume that under .ero load, a sin#le floc
ould $e spherical" %ts density is then >min"
%f e !no the force supported $y a sin#le floc,
e can calculate the total load of the refiner, as
e !no ho many flocs are in the #ap at each
time step" We no assume that a floc of radius D
responds to compression ith a hei#ht reduction
only, as illustrated $elo"
The force exerted $y the plates on the floc is
calculated in terms of the plate #ap #, as follos"
The contact area A $eteen the floc and the
plates is E <D@2-#@28F=, and therefore the density
of the floc is
>f 7m8E#<D@2-#@28+2=
0rom the Compressi$ility :/uation, e o$tain
the applied thrust on one floc in the #ap as

L 7 A <m8E#M<D@2-#@28+2==@<+8A=
here M,A are the constants #iven a$ove in the
compression e/uation"
Compression beteen bars
When the #ap $ecomes very narro, the flocs
cannot fit completely into the #ap" %f this is not
assumed, then it is found that the model
predictions are unrealistic, in that the motor load
continually increases" %n this situation e
suppose that the floc can partially enter the #ap
and $e compressed" The $asis of the calculation
is shon in the dia#ram"
The proportion of the floc in the #ap is denoted
$y p and is #iven as a function of #ap # and floc
diameter d $y the e/uation p7<#8d=@/, here / is
a constant" %n practice, the value assumed for /
has not much influence on the computed motor
load" The various curves presented a$ove for
motor load a#ainst #ap ere computed usin#
various values of /" This assumption is similar to
the common assumption that flocs are stapled
onto the leadin# ed#e of the $ars, and consistent
ith the o$servation that narro #aps lead to
fi$re cuttin#"
Calculation of the Motor !oad
At each time step, e calculate the dra# created
$y each floc on the plates, and add this up over
all flocs" Thus if e suppose that the same G
tan#ential coefficient of friction applies to each
floc, then the i-th floc contri$utes to the motor
load
Motor load from one floc 7 GL Hr ,
here H is the rotation rate and r is the radial
position of the floc <hich ill $e different for
each floc=" Durin# the simulation, a record is
!ept separately of the radial position and the
re#ion of each floc, therefore the computer
simulation can run throu#h the ta$le of flocs and
compute the contri$ution to the motor load from
each floc"
Implementation
A pro#ram as ritten in C to calculate the
properties of this model" Up to +9999 flocs could
$e simulated ithin the refiner ithout difficulty"
A ta$le is maintained durin# the simulation ith
one entry for each floc" As flocs enter the refiner,
they are added to the ta$le,and those leavin# are
removed" 0or each floc, its current radial
position, to#ether ith its re#ion and velocity are
recorded" 0rom this ta$le, the local densities
throu#hout the refiner can $e computed"
The physical parameters used ere
0loc si.e <mass=3 9"++ #
Minimum floc density3 +FI !#8m@4
Uncompressed floc radius3 J"I mm
Maximum density in any re#ion3 +999 !#8m@4"
The refiner parameters ere typical values3
%nner radius3 F4 cm
Outer radius3 JC cm"
Dotation speed3 +KK rad8s
0eed rate3 +"9J !#8s
Results and Discussion
The computed motor load as a function of plate
#ap has already $een presented in the
introduction" -ince feed rate is assumed constant,
the #raph of specific ener#y is similar" The order
of ma#nitude of the computed load is reasona$le
for a refiner, $ut the real interest lies in the shape
of the curve" %t has the same shape as the
/ualitative curve postulated $y *J, to descri$e the
pro$lem of plate clash" The main contri$ution of
this model has $een to demonstrate a mechanism
to explain this curve"
-ince this has $een a purely theoretical
investi#ation, the next step ould $e to compare
ith experiment" Unfortunately, since
commercial practice re/uires that plate clash $e
avoided, there is no /uantitative data availa$le in
the literature"
A shortcomin# of this investi#ation is the fact
that the simulation has $een conducted assumin#
that the operatin# conditions are steady"
Discussions ith mill operators su##est that the
main dan#er of plate clash does not arise durin#
steady operation, $ut rather durin# the period in
hich a refiner is $ein# moved from one
operatin# set point to another" The $asic
simulation method used here can handle time
dependent conditions, and some results ere
included in *4,, $ut they do not shed any
additional li#ht on plate clash"
0rom the theoretical point of vie, a reason for
pursuin# this or! is that plate clash is a
/ualitative phenomenon" Most models of refiner
operation contain many adLusta$le parameters" %f
the only test is /uantitiative a#reement ith
measurements, then it is difficult to test a theory
ith many adLusta$le parameters" The pro$lem
of plate clash, hoever, is a /ualitative effect that
should $e su$Lect to simulation and testin# hile
remainin# insensitive to actual parameters"
References"
*+, M") M%L:- AAD W"D MAN" The flo of
pulp in chip refiners" 1ournal of &ulp and &aper
-cience, +I<2=31I4-1C2, +BB9"
*2, O" 0AA, D"1" 1:00D:N, AAD D" OU:LL:T"
A stochastic model for the residence time of pulp
in a sin#le disc chip refiner" 1ournal of &ulp and
&aper -cience, +B<I=312IK-2CF, +BB4"
*4, 0" AL0OA-%" Modelin# the mechanics of a
ood chip refiner" M"-c" Thesis" University
Western Ontario 299+"
*F, -"T" 'AA" Compressi$ility and &ermea$ility
of 0i$re Mats" &ulp and &aper Ma#" Can" &p IJ-
CC, +BIB"
*J, P"A"DUMOAT, A"D"L:PAULT, 1"-"MACM,
1"'"DOP:D-, Computer Control of a TM&
&lant" &ulp and &aper Canada K4<K= T22FQ22B,
+BK2"

You might also like