Anda di halaman 1dari 21

1

Modelling Paper Structure and Paper-Press Interactions


Nikolas Provatas
1
and Tetsu Uesaka
2
1
Department of Materials Science and Engineering,
McMaster University, Hamilton, Ontario, Canada
2
Pulp and Paper Research Institute of Canada,
Pointe-Claire, Quebec, Canada
We introduce a new mathematical model of paper structure and paper-press interactions.
The model incorporates essential features describing paper structure and its development,
including fibre mechanics and stochastic fibre deposition, and most importantly, paper
compression occurring during pressing, calendering and printing. The power of
mathematical modelling as a predictive tool is demonstrated by two specific examples
related to paper structure development and print quality. The first example examines
effects of fibre furnish on surface structures, specifically identifying the causes of surface
pore formation which affects ink and coating holdout. The second example investigates
factors controlling the paper-plate contact during printing. We show that a combination
of compressibility and surface roughness parameters is required to quantify the contact
area.
INTRODUCTION
Print quality prediction is one of the most difficult problems for industry because of the
complex surface and bulk structures of paper and complex paper-press interactions. The
process of printing involves several fundamental steps, including contact of the paper
with a printing plate (or blanket in the case of offset printing), transfer of ink to the paper,
and the subsequent setting and drying of ink onto the paper [1]. The transfer of ink to the
paper web requires the print nip to come into proper contact with the paper at the
intended printed area(s). Poor paper-plate contact can be caused by a variety of
mechanisms, including heterogeneity in surface roughness, porosity, and local
compressibility, all of which can lead to a non-uniform ink-transfer and ink distribution.
Ink drying and penetration of ink solvent into the paper can similarly depend on surface
roughness, porosity, and local compressibility. The interplay of such structural factors of
paper with ink transfer and ink drying can lead to various print quality problems such as
missing dots, print mottle and set-off.
Identifying and measuring paper attributes (e.g., roughness, compressibility, fibre
furnish) controlling print quality are complex tasks. This is partly due to the large number
of structural and printing parameters involved in the print quality evaluation. Another
reason is that many of the fundamental paper parameters are either inaccessible to
measurement or too time-consuming and labour intensive to obtain. As a result, most of
the efforts to improve print quality have been largely based on empirical knowledge
obtained by a trial and error approach.
2
An approach that overcomes many of these problems is the use of computational and
mathematical modelling. Computational simulations based on mathematical models of
physical processes are capable of predicting and monitoring the complex interactions
between a number of parameters in processes involving paper structure [2,3,4,5]. They
are also able to overcome many traditional barriers imposed by laboratory experiments as
they provide full access to all information involving any physical process they are
playing out via simulation.
Many previous paper structure models [5] dealt with the bulk (strength, elastic
property, porosity, bulk, permeability) rather than local stochastic properties (surface
roughness, texture, local pore structure) that critically influence surface-plate contact and
other central issues in printing. The first successful stochastic model of paper structure
was the PAKAA model of Ref. [2]. Combining many features of fibre flexibility and
deposition that go into the development of a paper web, the PAKAA model was able to
capture some statistical features of paper [2,3], including bulk properties such as fluid
flow through paper [4].
This work presents the framework of a new stochastic fibre network model that extends
the PAKKA model of Ref. [2]. Our formulation incorporates local formation control,
distributed fibre orientations and fibre dimensions, non-uniform fibre collapsibility and,
most importantly, the effect of local paper consolidation and paper-press interactions.
Model parameters are formulated in terms of quantities that can in principle be measured
or controlled. These include fibre geometry, collapsibility and flexibility, the latter two
being related, via the fibres mechanical properties, to a phenomenological parameter
describing the overall force of fibre sedimentation during the forming process. This
model is expected to be the first step in a computational platform that will be used to
predict and diagnose paper quality issues and to develop effective papermaking
strategies. Its main difference from other paper structure models is that it simulates
details of paper microstructure across several length scales. These range from the micron
scale of individual pores or fibre crossing to centimetre-sized paper sheets. Such a multi-
scale model will enable us to link the local microstructure of paper to specific paper
performance properties.
This paper has two main parts. The first part summarises the general framework of
models theoretical and computational aspects. The second part demonstrates the
robustness of the model by applying it to two generic problems. The first examines the
formation of through-the-thickness holes in paper, while the second examines paper-plate
contact in printing. The first example is relevant to coating and ink holdout and the
second to speckle and missing dot formation in printing. The development of the
experimental methods and the validation of the model through these will be the subject of
subsequent reports. All mathematical details are given in the Appendices.
PART 1: THE MODEL
3D Fibre Network Model of Paper
3
This section describes the fibre network model. It describes simulated paper structure
prior to consolidation. As such it incorporates random fibre sedimentation, as well as the
mechanical features of fibre bending and collapse occurring during the initial stage of
paper forming. Details on mathematical formulae are presented in Appendix A. Paper
structure resulting from the consolidation of the paper web that occurs during wet
pressing and calendering is discussed in the next section.
Fibre deposition:
Paper structure is investigated using a fibre deposition model that simulates 3D fibre
webs comprising any combination of fibres, fillers and fines. The basic fibre web is
constructed by depositing, one at a time, particles (fibres, fillers or fines) onto a 3D-fibre
web situated in the y x plane. Particle centres are distributed randomly within a system
of dimensions
x x
S x S < < and
y y
S y S < < . Their angle of orientation is
distributed according to an angular distribution ) ( f . In the work presented here
is uniformly distributed in the range [-/2,/2], which corresponds to the case of
handsheets. Machine made papers can be easily
represented by anisotropic angular distribution functions [6]. Particles have a length L ,
width W and thickness H , where L , W and H are each derived from a distribution
function specific to fibre, filler or fines. Fibres also possess a cell wall thickness ,
defined implicitly through their coarseness, fibre dimensions and cell wall density. (For
cellulose this is approximately 1500 kg/m
3
.)
Fibre interactions:
In order to simulate the effect of fibre-fibre and fibre-web interactions, a particle
deposition rule is defined. A deposition event is always accepted if a fibre is deposited on
a section of the web whose mean height is S , where is the average thickness of
the paper over the projected area of the deposited fibre, is a constant in the range ] 1 , 0 [
and S is the average thickness of the entire paper. A deposition event not satisfying this
condition is accepted with a probability 1 0 < < , known as the acceptance probability
[7]. If the fibre is not accepted under these conditions, the deposition trial is repeated
until it is accepted. Particle deposition occurs until the required basis weight of the
simulated paper is attained. The vector ) , ( serves as a formation control parameter.
Through these parameters, the degree of fibre flocculation and mass density uniformity is
controlled in simulated paper. The choice 1 = produces uniformly random fibre
networks for all vales of . As 0 , the network can be made progressively more
uniform than that of a purely random deposition of fibres. This happens by favouring
deposition into valleys of the fibre web where the mean paper thickness is a fraction
of the average paper thickness. We note that in the limit 0 simulation times required
to produce a fibre network increase dramatically.
The simplest manner to deposit fibres is to distribute their centres and angular
orientations according to a uniformly random distribution function. It is well known that
even such simple deposition statistics capture many of the essential characteristics of
paper structure [6,8,9]. In real paper, however, complex formation hydrodynamics can
4
lead to formation structures that deviate from the uniform random deposition structure.
Simple effective interactions between fibres have been shown to produce long-range
correlation in the mass density of 2D simulated fibre networks [5,7]; an effect also found
in real paper [10]. A preliminary investigation using our paper structure model has shown
that fibre networks with a mean fibre length of 1mm can produce macro-flocs in the
range of 3-5 mm using formation parameters, 1 . 0 = and 1 = . Different floc sizes are
possible by varying the formation control parameters defined above. We should note that
using different types of fibre length distributions alone does not produce this large-scale
flocculation.
Fibre collapse and flexibility:
During deposition of the initial fibre web prior to consolidation, fibres undergo bending
and some collapse due to the draining force. These two processes are related to fibre
mechanical properties and a pressure exerted on the fibres in the forming section.
Deposited fibres are represented by a collapsed thickness
i
HC h = , which is a fraction of
their initial thickness. The degree of collapse is represented by
i
C that reflects the fibre
lumen collapse during deposition. We use a mathematical expression for
i
C similar in
form to that presented in Refs. [11,12]. We extend the original formalism for our model
by defining a dimensionless parameter f which, we assume, is proportional to a (uniform)
pressure acting on the top surface of all fibres during deposition. Through this definition,
i
C is expressed as
2
1
2
) ( exp 1 1

|
|
.
|

\
|
(

= f C
W
C
FA i

, (1)
where ) (
FA
C is a dimensionless function of the cell wall fibril angle
FA
, representing
the effect of the fibril angle on the transverse Youngs modulus of a fibre [12]. The
parameter f allows us to model the relative collapse of different fibres under similar
drainage conditions. (The specific values of ) (
FA
C and f are given in Appendix A.)
During deposition fibres also bend around structures on the developing paper surface,
denoted throughout by
) , ( y x S z = , (2)
where z is the surface height of the developing paper web. Fibre bending in the z-
direction is represented analogously to Ref. [2]. However, the procedure is modified to
take into account general fibre geometry. Also, fibre
surfaces are defined on a finite element mesh local to the fibre. Thus, the inter-fibre and
the fibre-web surface interactions during deposition are made using a finite element
interpolation [13] to avoid projection of fibres onto a discrete lattice. The bending
procedure is as follows. The fibre is first discretized in units of length scale . In this
report, m 5 = . We denote discrete positions along the fibre by ( j i, ), and the surface
5
heights of the bottom of a fibre by ) , ( j i z
B
. Each point of the fibre is pressed into the z-
direction as low as possible such that it does not move through the existing paper surface,
and such that any two nearest neighbouring points j i, and j i , on the fibre satisfy
f
B B
T
h
j i z j i z

) , ( ) , (
. (3)
The specific form of the flexibility parameter
f
T is given by
f
W
h
W
h
C f
T
FA
f
(
(

|
.
|

\
|

|
.
|

\
|

3
4
4
1 1 1
) (


, (4)
where f is a dimensionless geometric parameter related to the deviation of a typical
fibre in the network from a perfect beam element. Once determined, it is kept constant for
all simulated paper structures. (The value of f is given in Appendix A.)
Figure 1 illustrates an example of a simulated paper surface of a typical unconsolidated
fibre network. The units in the x-y axes are m 5 = , as is the case for all surface plots
presented in this paper. The colour scale represents height, with red denoting high points
on the surface and with blue denoting low points. Two generic species of fibres have
been used; one with mean length m L 500
1
= and the other with mean length
mm L 5 . 1
2
= . Their weight fractions were 30% and 70%, respectively. Both species had
their lengths drawn from a Poisson distribution, as were their widths. The mean fibre
widths were m W 30
1
= and m W 40
2
= , respectively. The coarseness of species 1 was
m mg/ 21 . 0 , while that of species 2 was . / 17 . 0 m mg Collapse indices used were 57 . 0
1
=
i
C
and 21 . 0 C i
2
= . Corresponding flexibility parameters were 26 . 0
1
=
f
T and 4
2
=
f
T ,
respectively. (Note superscripts denote species not exponents.) The formation control
parameters have been set to ( , ) (0.01, 0.94) = , while the parameter controlling force
on depositing fibres has been set to 0.017 f = (Appendix A). The basis weight was
2
/ 40 m g . The surface is generated by interpolating the top surface of individual fibres
into a 3D faceted surface function. The very rough surface structure shown in this figure
reflects the unconsolidated state of the network prior to any pressing. This choice of fibre
species is generic and merely serves to illustrate the models features.
Consolidation of Simulated Paper Structure
A Formalism for Compression of the paper Web:
Research using simulated fibre networks [2-9] has often met with limited success in
comparing theoretical predictions with experiments. This is because the models neglect
paper consolidation brought about by the processes of pressing and calendering. As a
6
result, fibres in simulated paper structures are unrealistically twisted and bent, and the
bulk and surface structures they predict are different from real paper structures.
Modelling compression of a paper begins by describing how the paper is deflected by a
pressing plate (press rolls, calender rolls, printing plate or blanket). This is modelled by a
general formalism based on ideas of elasticity theory, contact mechanics [14,15] and the
compression of disordered materials [16-20]. Its main details are given in Appendix B.
The formalism treats paper as a slab with a given surface profile and mass density
distribution. A fundamental assumption is that a stress concentration on the surface of the
paper cannot propagate its influence laterally beyond a correlation length scale . The
precise definition of this length scale is still under investigation. In this paper we assume
is the correlation length of the mass density. This assumption implies that we can
coarse-grain paper into independent mass springs of lateral size scale and whose
length is the average local thickness of the paper. The scale is assumed to be smaller
than or in the same order of the scale of certain printing defects such as mottle [19,20].
Algorithmically, this formalism for paper-plate contact is implemented by decomposing
the paper and pressing plate into a set of springs [15]. In this case the paper is locally
discretized into spring elements of size ( m 5 = , as for fibres). Each such
element has a local compressive tangent Youngs modulus ) , ( y x E

, which is a function
of the local mass density of the paper, coarse-grained over the length scale . The
pressing plate is similarly decomposed into a set of springs, whose stiffness is determined
by the pressing materials Youngs modulus. (The spring constants of the pressing plate
and the paper are defined in Eqs. (B6) and (B7) in Appendix B.) Contact between paper
and pressing plate is modelled by bringing the two spring fields into contact with one
another (see schematic in Fig. 2a). Allowing springs in each material to deform according
to local mechanical equilibrium (at locations where contact actually occurs) gives the
local stress distribution ) , ( y x P acting on the paper surface. The corresponding z-
direction strain created in each spring is transferred isotropically to the corresponding
fibre segments below the surface. Once all fibre co-ordinates have been accordingly
adjusted, a new surface function is generated. This compression model thus describes
surface and bulk deformation throughout the domain of the simulated paper. The
accuracy of the model is greatest beyond the correlation length , although it also yields
qualitative information below this scale. As illustrated in Fig. 2a, simulations presented
here consider only a one-side contact problem. Since the order of the correlation length is
on the scale of the paper caliper, a complete description requires two-sided compression.
This requires a modification of the spring-field model, and is currently under
development.
Computing the local Elastic Youngs modulus:
In this work, local paper compression is modelled with a piecewise elastic-plastic stress-
strain behaviour, followed by strain hardening, as typically seen in cellular foam
materials [17]. In the elastic range, the local elastic Youngs modulus ) , ( y x E

appearing
in the compression is assumed to depend on the local paper density coarse-grained on an
7
area . It has been found that the compressive modulus of a wide variety of cellular
and porous materials is described, to a high accuracy, by a function of their local density
[17,18,19] as
m
c
c
y x
E y x E
|
|
.
|

\
|
=

) , (
) , (
, (5)
where

is the local mass density,


c
the density of the cell wall and
c
E its Youngs
modulus of the cell wall. The values of m and
c
E depend on whether Eq. (5) expresses
an initial or incremental compression modulus. It was reported that for the initial modulus
of wood, , 3 m while 2 m in foams [17]. In the case of paper, there has been little
systematic data accumulated based on Eq. (5). An estimate of the incremental
compressive modulus gave 5 . 16 = m [23], while the initial compressive modulus seems to
scale linearly with paper thickness [24]. In this report we tentatively use the initial
compressive modulus estimated from Eq. (5) using 1 m and . 8MPa E
c
=
Plastic deformation of paper:
Consolidation during wet pressing (i.e., initial consolidation after fibre deposition)
involves plastic deformation of the paper structure, due to the plastic collapse and
buckling of individual fibre walls [17,18,19]. In this case the surface does not recover
much from the strain of initial compression [24,25]. Plastic compression of paper is
modelled with a non-linear stress-strain curve using the local tangent method [13].
Specifically, when the stress in a local paper spring exceeds a certain plastic collapse
stress ) , ( y x
PL
= , we set ) , ( ) , ( y x E F y x E
PL
PL

= , where 1 0 < <


PL
F and ) , ( y x E

is
the initial modulus determined from Eq. (5) using the local density of the paper web prior
to compression. The collapse stress itself is modelled by the relation
) , ( ) , ( y x E S y x
PL PL
= , where 1 0 < <
PL
S . The values of
PL
F are not known for paper
at this stage, although it is expected to be small as for most cellular materials [17]. We
thus tentatively set 1 . 0 =
PL
F . Similarly we expect the collapse stress to be a small
fraction of the initial modulus [17]. We set 0.15
PL
S = . (We note that the final sheet
structure details were relatively insensitive to a range of small values of
PL
F and
PL
S ).
Compression is applied in small increments, so that the local modulus is adjusted at each
time step to allow different regions of the simulated paper to enter the collapse regime
when their local stress exceeds ) , ( y x
PL
.
The compression formalism defined in Appendix B can be equivalently applied when
consolidation pressure is removed from the pressed paper web. In regions where plastic
collapse occurred during initial consolidation, partial strain recovery will occur. This is
modelled by a linear stress-strain relationship whose slope (i.e., local recovery modulus)
is set to a factor
RE
F of the initial modulus below the collapse stress [13]. The factor
RE
F
increases with larger consolidation loads, since the more fibres are collapsed the less they
recover. The recovery factor is not known for paper. We illustrate simulated
8
consolidation with 2 =
RE
F for webs consolidated to a thickness greater than m 85 and
3 =
RE
F for a consolidation thickness below this.
Subsequent to the initial consolidation (and partial recovery) described above, paper
becomes strain-hardened. Since most of the fibres are completely collapsed, they will
exert considerable pressure upon further application of pressure (such as further pressing
during printing, as in the application of the model presented in Part II of this paper).
Compression in this regime is thus assumed to occur elastically [17], with a strain-
hardened local modulus given by Eq. (5) re-assigned to the local fibre springs of the
network. In this case the higher density of consolidated sheets gives rise to a much higher
modulus in this phase.
Consolidation of paper bulk:
It was noted above that strains induced in the surface are transferred to the fibres within
the bulk. This is achieved by re-packing the fibres of the fibre network at each
incremental step of the consolidation process. There are three steps in this re-packing
process: (1) Obtain surface strains according to deflected springs of the spring-field
model; (2) use local surface strains to homogeneously deform fibres in the bulk; and (3)
update fibre networks surface topography according to the adjusted fibre packing.
Following step (3), the above procedure is repeated until a desired compression pressure
or average paper caliper is reached. The assumption that surface strain at a location in the
x-y plane is transferred homogeneously to all fibres is a reasonable assumption that has
been used in other work to predict bulk compression of fibre assemblies [27].
Figure 2b shows a flow chart illustrating the various steps of the fibre network model
described above. In this paper, the physical constants
PL RE PL c
S F F E
,
, , were treated as
phenomenological parameters, set to values that yielded consolidated paper structures
with target callipers and surface roughness corresponding to typical commercial
newsprint. Their precise quantification is currently under investigation for paper using a
new continuum mechanics investigation, which is described in more detail at the end of
Appendix B
Figures 3a and 3b show an example of a simulated paper web before and after
consolidation. Two generic fibre types (type A and B) have been used. Their mean
lengths are mm L
A
5 . 1 = and 500
B
L m = , respectively, while their corresponding mean
widths are m W
A
40 = and 30
B
W m = . The coarseness values of the two fractions are
m mg/ 168 . 0 and m mg/ 190 . 0 , respectively. Fibre flexibility values are 4 =
A
f
T and
35 . 0 =
B
f
T , while their collapsibility values are 21 . 0 =
A
i
C and . 53 . 0 =
B
i
C The weight
fractions of species A and B are 35% and 65%, respectively. The formation parameters
during deposition are set to ) 94 . 0 , 001 . 0 ( ) , ( = , and deposition force parameter is
017 . 0 = f . Figure 3a shows the surface of the paper before consolidation, revealing large
variations in surface topography. Consolidation of the fibre network produces an even
9
caliper and a smoother surface (Fig. 3b). The fibres in the bulk have been accordingly
consolidated.
PART 2: APPLICATIONS
Paper Structure and Surface Holes
Ink holdout and coating holdout are influenced by the presence of excessively large
surface pores. Their creation appears to be closely linked to the fibre furnish as well as
formation. However, the relationships between surface pores and such papermaking
factors are not known. In order to understand the possible causes leading to the creation
of large surface pores, we simulated paper structures using two types of generic fibre
species: one collapsible and flexible fibre (Furnish A), and the other, a very stiff and non-
collapsible fibre (Furnish B).
Figure 4a shows the typical surface of paper made by using only Furnish A. The
dimensions of fibres simulated are m L 500 = and m W 50 = . Collapsibility and
flexibility are 2 . 0 =
i
C and 2 =
f
T , respectively. The formation control parameter was
set to ) 1 , 1 . 0 ( ) , ( = . Fibre orientations were assumed to be isotropic. The deposited
fibre web (i.e., the unconsolidated sheet) was consolidated using the compression model
described earlier. While the surface roughness of the consolidated paper was fairly high
(approximately m 6 ), it is clear that there are no large surface pores extending into the
bulk. Figure 4c shows the pore-size distribution corresponding to this paper. There is one
main peak, which is typical of these types of papers.
The addition of even a small fraction of stiffer, coarser fibres makes a large difference in
the porosity and surface characteristics of the paper. Figure 4b shows a simulated paper
surface with conditions similar to those of Fig. 4a, but with approximately 8% (by
weight) coarser, Furnish B, fibres added. The dimensions of the additional Furnish B
fibres are mm L 1 = and m W 50 = . The collapsibility and flexibility of Furnish B
are 1 =
i
C and 26 . 0 =
f
T , respectively. Formation was now set by the parameters
) 94 . 0 , 01 . 0 ( ) , ( = . These formation parameters were chosen because they improved
formation (in terms of the coefficient of variation of mass density) as compared to the
paper simulated in Fig. 4a. The overall surface roughness was m 3 . The remarkable
feature of this simulated paper surface is the appearance of large surface pores, or pits,
extending deep into the bulk of the paper (Fig. 4b). Their lateral dimensions could exceed
m 50 . Figure 4c shows a corresponding pore-size distribution. There is clear emergence
of a shoulder in the range of 20-100 m that is at least an order of magnitude greater than
the shoulder in pore size distribution of corresponding to Fig. 4a. We note that simulating
the network in Figure 4b with the same formation parameters as in Figure 4a also leads to
the formation of large surface pores.
10
These result suggests that even a small fraction of non-collapsible, stiff fibres present in
the furnish may create large surface pores. This can have negative consequences on ink
and coating holdout properties. This emergence of surface-holes did not change
significantly by altering the mean fibre length or width of either species. The existence of
a small fraction of the less flexible fibres suffices to create large surface-holes.
Surface Structures and Paper-Plate Contact
In this section, we have applied the fibre network model to examine the role of surface
and bulk properties on the paper-plate contact. The specific situation we model is that of
hard plate-paper interactions, which may correspond to the case of gravure and letterpress
printing. An advantage of simulating paper-plate interaction is the ability to model the
contact process in-situ. Figure 5 illustrates a snap-shot of a simulated paper surface
under the compression of a printing nip. Indicated are typical contact and non-contact
areas. It is clear that different areas of the paper can compress differently, leading to non-
uniform paper-plate contact.
Figure 6 plots the computed non-contact areas of two simulated paper surfaces under the
pressure of a hard-plate print-nip. The difference between the two papers is in their initial
average surface roughness and curvature , compactly defined through a
dimensionless measure or texture, , often used to describe a random surface [15].
Both paper webs were initially formed with the following parameters: Formation
parameters ) 94 . 0 , 01 . 0 ( ) , ( = , and the parameter controlling force on depositing fibres
0.017 f = . Two generic species of fibres similar to those in the previous examples were
used: Species A (short fibres) and Species B (long fibres). The mean length of Species A
is 250 m and for Species B, 1.5mm. The mean widths of species A and B are 30 m
and 40 m , respectively. Their lengths are drawn from a Poisson distribution. The
coarseness of Species A is m mg / 21 . 0 , giving a collapsibility index of 57 . 0 C
A
i
= , while
for Species B, the coarseness is m mg/ 17 . 0 , giving 21 . 0 =
B
i
C . The corresponding
flexibility parameters are taken as 26 . 0 =
A
f
T and 4 T
B
f
= , respectively. The paper webs
comprised 30% of Species A, and 70% Species B (by weight). After initial fibre
deposition, the paper webs were consolidated using the compression algorithm described
in Part I. Since this phase corresponded to plastic collapse of most fibres (implemented
by using the non-linear stress-strain relationships described above), the sheet thickness
was reduced to over 60% of its initial value. The different levels of consolidation were
achieved by consolidating the fibre mats to different final heights (squeezed to different
final levels). Due to the plastic deformation, this generally left the network squeezed to
smaller thickness with a smoother surface. Following consolidation, subsequent
compression due to printing was assumed to take place within the elastic limit.
The main result inferred from Fig. 6 is that at a given pressure, the smoother paper with
less compressible bulk structure can actually have a larger non-contact area than the
rougher, more compressible paper. In the printing process where the nip pressure is
controlled, this result may have an important implication on the use of calendering to
11
improve paper-hard plate contact. While calendering increases smoothness, it also
reduces compressibility. Therefore, depending on the relative magnitudes of these
competing effects, calendering can improve or reduce paper-hard plate contact. A
negative consequence of these competing effects is, so-called, missing-dots of highly
calendered grades for gravure printing. (In the printing process where the nip
displacement or squeeze is controlled, the role of paper compressibility will be different.)
We can understand the data of Fig. 6 by deriving a mathematical relationship between
paper-plate contact area, surface structures and compressibility. We assume that the
change in the contact area
r
dA decreases in proportion to the amount of existing non-
contact area, and increases with an increase in the total printing pressure dP. This
hypothesis can be mathematically expressed as
dP A dA
r r
) 1 ( = , (7)
where is a proportionality constant assumed to depend only on initial paper structure
parameters, i.e., ) , , ( E = , at least within a range of pressures. The
modulus E in denotes an average over the entire printing area. The solution of Eq. (7) is
) ) , , ( exp( 1 P E A
r
= . (8)
Paper-plate contact has been found to follow Eq. (8) to a good approximation over a
range of pressures. As seen in Fig. 7, a plot of ) 1 ln(
r
A vs. P is approximately linear at
lower pressures, showing the validity of Eq. (8). (The parameter was determined from
the plot of ln(1-A
r
) vs. P for different simulated papers.) Note that the form of Eq. (8)
remains valid at higher pressures, however is different for each paper.
An explicit form of as a function of , , and E is not known. However, an insight of
the functional form of for low pressures can be obtained from Hertz theory [15,28],
Which gives ( ) E P A
r
/ / = . Expanding Eq. (8) at low pressures, we obtain
( , , )
r
A E P = (9)
Comparing Eq. (9) to the Hertz theory prediction gives
( , , ) E
E

= (10)

This Hertz model (linear contact model) derives the expression for contact areas of
surface peaks using equations for the compression of two spheres. It also assumes
compression at low pressures and a homogeneous, isotropic surface whose peaks have a
constant curvature, and whose height follows a Gaussian distribution. Therefore, we do
12
not expect this proportionality constant to be the same as the one in our model, which is
geometrically compressing mass springs. Figure 8 shows a plot of , which was
obtained from our simulation model, against 1/ E . Parameters E, , and were also
computed from the simulation model. The result shows that is surprisingly well
approximated by E / 5 . 74 , except at low values of 1/ E (less compressible and
rough sheets). We find that this behaviour is valid for bulky paper webs comprising
very flexible fibres, and of fairly even formation. Such bulky sheets are expected to
produce surface conditions that follow the conditions of Hertz theory. (Figure 1 is typical
of such surface geometry). These results show that in order to calculate the quality of
paper-plate contact for a given paper sheet, we need to know not only its surface
roughness and curvature but also its local compressibility.
CONCLUSION
This paper has presented a mathematical model to predict paper surface structures and
paper-press interactions in printing. The simulation model captures many of the important
effects of the papermaking process on the final paper structures. In particular, it
incorporates fibre orientations, wet fibre flexibility, fibre collapse, the distributions of
fibre dimensions, fibre-fibre interactions during forming, and most importantly sheet
consolidation by pressing and calendering. The fibre network model predicts surface and
bulk structures comparable to those found in commercial papers. This feature also allows
us to model paper-printing plate interactions under nip impression.
We demonstrated the predictive power of the model by applying it to two special cases.
The first example illustrated a mechanism behind the formation of surface pores and
demonstrated that a small fraction of coarse, stiff fibres has a great effect on the surface
pore formation. The second example examined the effect of surface compressibility,
surface roughness and local curvature on the paper-plate contact during printing. We
found that surface compressibility may be more important than surface roughness and
surface curvature to describe the paper-plate contact. We also developed a
phenomenological equation to describe paper-plate contact.
REFERENCES
1. ASPLER, J., Proceedings of the 2
nd
International Symposium on Printing and Coating
Technology, University of Wales, Swansea, p.1, Sept. 2000.
2. NISKANEN, K. and ALAVA, M., Phys. Rev. Lett.(73): 3475 (1994).
3. NILSEN, N., ZABIHIAN, N. and NISKANEN, K., A New Tool for Simulating
Paper Properties, Tappi J. 81(5): 163-166 (1998)
4. RASI, M., KOPONEN, A., AALTOSALMI, U. , TIMONEN, J. , KATAJA, M. and
NISKANEN, K., "Permeability of paper: Experiments and numerical simulations",
TAPPI International Paper Physics Conference, pp. 297-306, Sept., 1999, San Diego,
CA
13
5. SILVIE, J., CARET, C., BELAMAALEM, B., and MAHROUS, M., "The three-
dimensional structure of paper: Methods of analysis and implications on its physical
properties", 1995 International Paper Physics Conference, Niagara-on-the-Lake,
Canada, pp. 91-95.
6. DODSON, C.T.J., Paper, an Engineered Stochastic Process, TAPPI Press, Atlanta
(1994).
7. PROVATAS, N., HAATAJA, M., ASIKAINEN, J., MAJIANIEMI, S., ALAVA, M.
and ALA-NISSILA, T., J. Col. and Surf.(165): 209 (2000).
8. NORMAN, B. and WAHREN, D., Svensk Papperstidn., 77(11): 397 (1974).
9. CORTE, H. and KALLMES, O.J., Statistical Geometry of A Fibrous Network, in
Formation and Structure of Paper, Vol. 1, p. 13 (edited by F. Bolam), British Paper
and Board Makers Assoc., London (1962).
10. PROVATAS, N., ALAVA M., and ALA-NISSILA, T., Phys. Rev. E.(54): R36
(1996).
11. JANG, H.F., WEIGEL, G., SETH, R.S., and WU, C.B., Pap. Puu, 84(2):112-115
(2002).
12. JANG, H.F., The Science of Papermaking, Trans. 12
th
Fundamental Res. Symp.,
Oxford, UK, 1:193-210 (2001).
13. COOK, R., MALKUS, D.S., and PLESHA, M.E., Concepts and Applications of
Finite Element Analysis, John Wiley & Sons, New York (1989).
14. LANDAU, L.D. and LIFSHITZ, E.M., Theory of Elasticity, 3
rd
Edition,
Butterworth-Heinemann, Woburn, MA. (1959).
15. JOHNSON, K.H., Contact Mechanics, Cambridge University Press, Cambridge,
UK (1985).
16. RODAL, J.J.A., Tappi, 76(12): 63 (1983).
17. GIBSON, L.J. and ASHBY, M.F., Cellular Solids, Structure and Properties,
Pergamon Press, Oxford, UK (1988).
18. GARBOCZI, E.J. and DAY, A.R., J. Mech. Phys. Solids (43): 1349 (1995).
19. ROBERTS, A.P. and GARBOCZI, E.J., Los Alamos Condensed Matter Physics,
Preprint Archive, Number 0009004 (2000).
20. CHAIKIN, P.M. and LUBENSKY, T.C., Principles of Condensed Matter Physics,
Cambridge University Press, Cambridge, UK (1995).
21. KENNEY, M.C., KONOPASEK, L. and DENT, J.R., JPPS, 13(3): J82 (1987).
22. KAJANTO, I.M., JPPS, 17(5), J178 (1991).
23. OSAKI, S. and FUJII, Y., Japan Tappi, 32(8): 476 (1978).
24. YAMAUCHI, T., Appita, 42(3): 222 (1989).
25. GRATTON, M.F., 3-D Web Deformations from Calendering in a Hard Nip, 1996
Finishing and Converting Conference, Bal Harbor, Fl., USA, 173-180.
26. AMIRI, R., GORRES, J., GRONDIN, M. and WOOD, J.R., Products of
Papermaking, Trans. 10
th
Fundamental Res. Symp., Oxford, UK, 1:285-310 (1993).
27. KOMORI, T., ITOH, M. and TAKAKU, T., Textile. Res. J., 62(10): 567 (1992).
28. GREENWOOD, J.A. AND WILLIAMSON, J.B.P., Contact of Nominally Flat
Surfaces, Proc. Roy. Soc. London, Series A, A295: 300 (1966).
14
APPENDIX A
This appendix derives the various formulae presented in the text, and at the end indicates
the value of the various constants used.
Collapse equation:
Reference [10] derives an expression for the collapse index of kraft fibres as a function of
their cell wall fibril angle
FA
, under the application of a fixed pressure F . The collapse
index is defined as
o I
A A C / 1 = where Aand
o
A are the uncollapsed and collapsed
lumen areas, respectively. Its form is given by
2
2 2
1 exp ( ) ( ) ,
2
I FA
W H
C kF C

| |
+ | |
= |
|
|
\ .
\ .
(A1)
where is a constant and the combination ) (
FA
kC is the inverse of the transverse
Youngs modulus of a fibre (Eq. (C1) in Ref. [12]). The function ) (
FA
C is a
dimensionless parameter, representing the effect of fibril angle on the transverse Youngs
modulus. The constant k is elastic compliance (the inverse of the transverse Youngs
modulus) of a fibre when the fibril angle is equal to zero [12]. It depends only on fibre
wall material properties and is thus treated as a constant throughout. ]. The parameter F is
the pressure applied on fibres during drainage.
It is reasonable to assume that Eq. (A1) can be applied for a range of sedimentation
pressures F . We can thus replace the factor kF by one parameter f , which represents
the phenomenological pressure parameter acting on fibres during deposition.
Furthermore, by expressing the change in fibre height and width by r W W + = and
r H h = , respectively (this preserves the perimeter of a collapsed fibre), it is
straightforward to obtain
i
C in Eq. (1) from Eq. (A1). Note that the factor of in Eq.
(A1) was absorbed into f giving / f f = in Eq. (1). Note also that in deriving Eq. (1)
we assumed that H W = .
Fibre bending and flexibility:
To describe fibre bending during deposition, we let ( z , , ) denote the reference frame
fixed with the fibre and ( z y x , , ) denote the global reference frame. The former is an
internal reference frame that defines the lateral positions of the fibre on the rectangle
, 2 / 2 / L L < < and 2 / 2 / W W < < . Points local to the fibre frame are mapped
onto the global frame by a generalised transformation:
x=
x
T (
,
) and y=
y
T (
,
). (A2)
These transformations can incorporate rotation, translation, expansion and curl. The top
(T) and bottom (B) surfaces of each fibre (the same in either reference frame) are defined
through the functions
15
) , (
B T
F z = and
i B B
HC F z + = ) , ( . (A3)
During deposition, fibre bending stops when the distance between the bottom of the fibre
and the paper surface:
)) , ( T ), , ( S(T - ) , ( ) , (
y x

B
F D = (A4)
is minimised for all ) , ( , subject to the bending constraint:
f
B B
T
h
F F

) , ( ) , (
, (A5)
where ) , ( and ) , ( are any two distinct points on the fibre surface. This procedure
defines a functional minimisation for the function ). , (
B
F The constant
f
T is a
dimensionless flexibility parameter similar in physical meaning to that in Ref. [2].
Since we are simulating fibre deposition computationally, both (x, y) and (
,
) co-
ordinates are made discrete to a spatial resolution . In this case the bending criteria of
Eqs. (A4) and (A5) apply between discrete points ) , ( j i and ( j i , ), giving Eq. (3) in Part
1. Matching positions between the discrete internal fibre co-ordinates and continuous
global co-ordinates is done using linear isoparametric finite-element interpolation
functions [13].
The flexibility parameter
f
T in Eq. (A5) is approximated by considering a small portion
of fibre of length , cross section h W and cell wall thickness bent under a constant
force per unit length P acting uniformly on the top surface of the fibre. Specifically, we
consider the situation where the two ends of the length are freely supported. From
beam theory [14] the maximum deflection of such a beam is given by
L
IE
P
z
4
384
5
= , (A6)
where
L
E is the Youngs modulus of a fibre in the longitudinal direction and I is the
moment of inertia of the beam section, given approximately by
(
(

|
.
|

\
|
|
.
|

\
|
=
3
3
1 1 1
12 W
h
W
Wh
I

.
(A7)
The factor of 5/384 in Eq. (A6) will change depending on the boundary condition to
which the fibre segment is subjected during bending. The actual fibre bending is
geometrically much more complex than that described above. We can only expect that
the bending deflection is proportional to Eq. (A6). We therefore replace Eq. (A6) by
16
L
IE
P
f z
4

= , (A8)
where f is a free dimensionless parameter which can, in principle, be used to fit the
model fibre flexibility with experimental fibre flexibility measurements.
We assume that the pressure of collapse is the same as the pressure applied during
bending. To relate the bending pressure to the calibration pressure F used in defining
collapse above, we again use the definition f kF = and express
L
E P/ in Eq. (A8) as
L L L
kE
W f
W
kE
kF
E
P

= =
) (
, (A9)
where multiplication by the fibre width W is made since F is a pressure, while P is a
force per unit length. We further assume that the longitudinal modulus
L
E and the
transverse modulus
T
E are related via
L T
E ME = , where M is generally dependent on the
fibril angle and the elastic anisotropy of the fibril. As we have noted above, the inverse of
the transverse fibre modulus is given by Eq. (C1) of Ref. [12],
1
( )
FA
T
kC
E
= , (A10)
where ) (
FA
C and k were defined earlier. Using Eq. (A10) in Eq. (A9) (along with the
assumption
L T
E ME = ) we arrive at
M
f
W C
E
P
FA
L

) ( = , (A11)
Substituting Eq. (A11) into Eq. (A8), and recalling the definition from above f f = , we
obtain
f
W
h
W
h
C f
T
h
z
FA
f
(
(

|
.
|

\
|
|
.
|

\
|

3
4
4
1 1 1
) (


, (A12)
where ) /( 12 M f f = . As with collapsibility, Eq. (A12) allows us to describe fibre
flexibility as a function of fibre properties and the overall pressure acting on the paper
web during fibre deposition.
Parameters used in simulations:
17
In this work, the constants f and f were taken as phenomenological parameters set to
1 = f 00 and 017 . 0 = f , respectively. The constant , 5 . 0 ) ( =
FA
C as determined from
Ref. [10]. The choices for f and f were made such that the calliper of our fibre
networks (prior to pressing), as well as the degree of fibre bending and collapse was
consistent with typical newsprint grades. We note that equations (1) and (4) in the text
can in be used to accurately determine f and f by simultaneously measuring fibre
collapse and fibre flexure using atomic force microscopy technique.
APPENDIX B
Consolidation of paper:
To motivate our formalism to describe paper compression, it is instructive to first
consider the simple case of half-space of some isotropic elastic material bounded from
the 0 = z plane. An applied pressure ( , )
j
P x y will create a deformation
i
u of the plane
given by the expression [14]
y d x d y x P z y y x x G u
z
j ij i
=

=0
) , ( ) , , ( , (B1)
where ( , , )
ij
G x y z is the Greens function specific to this geometry [14,15]. The Greens
function is a propagator of the stresses ) , ( y x P
j
from the point ) , ( y x to the point
) , , ( z y x in the half-space. In the case of an isotropic body and a normal force, surface
deformation in the z-direction is given by the corresponding Greens function
( ) ( )
2 2
2
33
1
) , (
y y x x E
y y x x G
+

=


, (B2)
where E is the Youngs modulus of the space and is its Poisson ratio. Considering a
normal pressure P along the z-direction, we obtain the deformation of points on the
surface of the half-plane by substituting Eq. (B2) into Eq. (B1), obtaining [14,15]

=

+

=
0
2 2
2
) ( ) (
) , ( ) 1 (
) , (
z
z
y d x d
y y x x
y x P
E
y x u

. (B3)
We can also express Eq. (B2) with the observation point ( y x, ) as the centre of co-
ordinates. In doing so we define the co-ordinates of any other point ) , ( y x relative to
( y x, ) by
2 2
) ( ) ( y y x x r + = and )) /( ) arctan(( x x y y = (the angle subtended
by the line ) , ( y x - ) , ( y x from a reference axis). The total displacement at ( y x, ) is then
given by
18

drd r P
E
y x u
z
z

=

=
0
2
) , (
1
) , ( . (B4)
In the case of a constant pressure,
c
P y x P = ) , ( , applied over a circular area of radius R ,
Eq. (B4) shows that
z
u is of significant magnitude even at distances quite large compared
to the extent of the pressure field R [14,15]. This is a characteristic of the long-ranged
nature of elastic forces and has important consequences in materials such as metals or
incompressible materials.
The structure of paper is clearly quite different from many solid elastic or viscoelastic
materials [26]. In compression it behaves locally very much like a cellular material [17].
The highly disordered state of its constituent fibres, as well as the empty space within
fibres serves to damp stress propagation in the x-y direction and limit its influence to
distances quite close to the point of stress at the surface of the paper. Furthermore, paper
structure is anisotropic and is of finite thickness, not a semi-infinite plane.
The simplest model to limit the lateral influence of a concentrated stress at a point
) , ( y x is by defining a short-range cut-off beyond a distance (which we term the
elastic correlation length). In this report, we assume that this length scale is taken as the
decay range of the two-point mass-density correlation function [20], defined by

+ =
A
dxdy r x m x m r G ) ( ) ( ) (
r r r r
, (B5)
where ) (x m
r
represents the mass density of our simulated paper webs. (For a discrete
mass density the integral can be replaced by a summation in the x and y directions.) For
isotropic conditions, |). (| ) ( r G r G
r r
= For anisotropic conditions ) (r G
r
may generally have
different decay lengths
MD CD
and in the machine and cross directions, respectively. In
this report we present simulated paper structures with isotropic fibre orientations and
forming conditions. We thus consider the correlation length only in one direction.
The existence of such a cut-off conceptually allows the paper to be coarse-grained [20]
into independent spring elements of lateral size (
CD MD
for an anisotropic
case). Within each of these mass-springs, we define an effective compressive Youngs
modulus ) , ( y x E E

= where the co-ordinate ) , ( y x is the centre of the mass-spring. The
following important assumptions are made in using this decomposition: (a) is large
enough so that an effective local compressive modulus can be defined about the point
) , ( y x and (b) is small enough that lateral spatial variations in the paper, including
paper anisotropy, can be realised via ) , ( y x E

. These are the main assumptions of this


section.
Computationally, contact between paper and pressing plate (press rolls, calender rolls,
printing plates or blankets) can be incorporated into the simulation algorithm by
decomposing the paper into a Winkler foundation mattress [15]. Specifically, the discrete
19
elements of size comprising the paper surface are treated as independent springs of
compressive modulus ) , ( y x E

(a function of the local mass-density, coarse-grained


about the elements centre ) , ( y x ). The pressing surface is similarly decomposed into the
foundation mattress, with the stiffness of its elements determined by its material
properties. Paper-plate interactions occur by bringing the spring-fields of the paper and
pressing plate into contact (see Fig. 2). Mass-springs between which contact occurs
deform according to local mechanical equilibrium. This defines a local stress
distribution ) , ( y x P throughout the paper and pressing plate. Knowledge of the pressure
field ) , ( y x P allows us to compute surface and bulk deflections in the fibres of the
underlying fibre network. We note that strictly speaking, this algorithm becomes
physically correct only at length scales > r . On length scales below , the model is
only algorithmically practical, providing only qualitative trends of surface deformation
are described.
Explicit calculation of the pressure field ) , ( y x P proceeds as follows: We assign to each
spring of the pressing plate a stiffness constant
b
b
b
t
y x E
y x k
2
) , (
) , (

= , (B6)
where ) , ( y x E
b
is the z-direction Youngs modulus of the compressing plate and
b
t is its
thickness. The modulus of this plate depends on the printing process (i.e., low
b
E for
offset printing, and very high for gravure and letterpress printing). In this report we set
b
k to a constant over the whole pressing plate (uniform compressibility of the plate). We
similarly assign the paper mass-springs a stiffness
) , (
) , (
) , (
2
y x S
y x E
y x k
p

=

, (B7)
where ) , ( y x E

is the tangent modulus of a local paper spring and ) , ( y x S is the local


paper surface height before deformation. The co-ordinates ( y x, ) define the centre of a
discrete element.
Starting with the pressing surface and the paper out of contact, we incrementally lower a
reference point
R
y on the pressing surface in steps of z toward the paper surface. Each
increment can be regarded as a pseudo time step n , during which
R
y is lowered from
some arbitrary initial position
o
z to z n z
o
. At points ( y x, ) where the two surfaces
come into contact (see illustration in Fig. 2), the springs of either medium deflect in
accordance to local mechanical equilibrium. Let ) , ( y x R
n
and ) , ( y x S
n
denote the
heights of the plate and paper surfaces at a time step n . At points where contact has
20
already been made ) , ( ) , ( y x S y x R
n n
= . The heights at time 1 + n are found as follows. At
positions where ) , ( ) , ( y x S z y x R
n n
> (no contact)
1
1
( , ) ( , )
( , ) ( , ).
n n
n n
R x y R x y z
and
S x y S x y
+
+
=
=
(B8)
At positions where ) , ( ) , ( y x S z y x R
n n
< , contact equilibrium between two springs
requires that their forces are equal and opposite. This is written as
{ }
{ } ) , ( ) , ( ) , (
)) , ( ( )) , ( ( ) , (
1
1
y x S y x S y x k
y x R y y x S z y y x k
n n p
n R n R b

=
+
+
(B9)
The solution of Eq. (B9) gives
b p
n b n p
n n
k y x k
z y x R k y x S y x k
y x R y x S
+
+
= =
+ +
) , (
) ) , ( ( ) , ( ) , (
) , ( ) , (
1 1
. (B10)
From the deflections defined in Eq. (B10), we compute the pressure increment field
acting on the paper by
) , (
) , (
) , ( ) , (
) , (
1
1
y x E
y x S
y x S y x S
y x P
n n
n
|
|
.
|

\
|
=
+
+
. (B11)
The procedure defined above is incremental. Thus, it can be applied to any non-linear
stress-strain path (such as that occurring during plastic collapse or partial recovery) by
adjusting the local moduli of the network, ) , ( y x E

and
) , ( y x E
b
incrementally.
Several important points need to be made regarding this model. As mentioned above, this
correlated mass-spring model is accurate on length scales beyond the correlation length.
We have found that the correlation length is typically of order m
2
10 . This length
scale is still sufficiently small to resolve a detailed compressibility landscape which can
display structure on millimetre length scales [20]. This is also small enough to resolve
some macro-scale printing defects, such as mottle, which can display millimetre-size
features [21]. For the case of micro-scale print defects, such as missing dots, of size in the
order of 100 m or smaller, this simple model can only be used qualitatively.
Determination of elastic properties of paper:
The physical parameters
PL EL PL c
S F F E
,
, , defined in the compression model are being
investigated using a new continuum phase-field model of deformation. This model
treats any generalised geometric assembly of fibres in a domain as a continuum field with
phase-dependent elastic constants. Through knowledge of the elastic or plastic properties
21
of the cell wall material of local constituent fibre, the model self-consistently solves for
the exact distribution of stresses and strains within the domain of the fibre web being
examined. By transferring these strains to the corresponding finite elements of individual
fibres, we will be able to predict the web deformation under tensile of compressive loads.
This information can then be used to determine the constants defined above
Figure Captions:
Figure 1: Simulated paper surface. The colour code represents height, with light colours
denoting higher points and dark colours lower points. A mixture of two fibre furnishes
was used to construct this surface, as described in the main text. The horizontal axes are
in units of m 5 .
Figure 2: (a) Spring-field model. The paper-plate interaction is modelled by bringing
commensurate springs from each field (paper and plate) into contact. (b) Flow chart
illustrating the steps of the paper structure model algorithm.
Figure 3: Simulated paper before (a) and after consolidation (b). The final caliper and
roughness of the sheet were approximately m 70 and m 5 . 4 , respectively.
Figure 4: (a) Simulated paper surface composed of almost fully collapsed fibres of one
type of furnish. (b) Paper constructed with a large fraction of flexible fibres and a small
fraction of very stiff, non-collapsed, coarse fibres. (c) Pore-size distributions
corresponding to Fig 4(a) (bottom curve) and Fig. 4(b) (top curve). The top curve shows
the emergence of many large pores on scales greater than 20 m . The horizontal axis is
in units of . 5 m
Figure 5: A snapshot of the paper surface during the process of compression of a
printing plate calculated using the compression model.
Figure 6: The non-contact areas of simulated papers under printing pressure. The two
papers differ only in the level of their initial consolidation. The less compressible paper
has the smoother surface, yet makes poorer contact with the plate.
Figure 7: The non-contact areas for a wide range of simulated paper samples as a
function of printing pressure. The data is described by the exponential function of Eq. (8)
over a wide range of pressures.
Figure 8: The fitted values of 013 . 0 from the data of Fig. 11 plotted against 1 E k
in Eq. (10).

Anda mungkin juga menyukai