19.01.2015 Views

3D network simulations of paper structure - Innventia.com

3D network simulations of paper structure - Innventia.com

3D network simulations of paper structure - Innventia.com

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

PAPER PHYSICS<br />

<strong>3D</strong> <strong>network</strong> <strong>simulations</strong> <strong>of</strong> <strong>paper</strong> <strong>structure</strong><br />

S. Lavrykov, S. B. Lindström, K. M. Singh, B. V. Ramarao<br />

KEYWORDS: Fiber Network, Random, Hand sheet,<br />

Bending Stiffness, Elasticity, Compression<br />

SUMMARY: The <strong>structure</strong> <strong>of</strong> <strong>paper</strong> influences its<br />

properties and <strong>simulations</strong> <strong>of</strong> it are necessary to<br />

understand the impact <strong>of</strong> fiber and <strong>paper</strong>making<br />

conditions on the sheet properties.<br />

We show a method to develop a representative <strong>structure</strong><br />

<strong>of</strong> <strong>paper</strong> by merging different simulation techniques for<br />

the forming section and the pressing operation. The<br />

simulation follows the bending and drape <strong>of</strong> fibers over<br />

one another in the final <strong>structure</strong> and allows estimation <strong>of</strong><br />

sheet properties without recourse to arbitrary bending<br />

rules or experimental measurements <strong>of</strong> density and/or<br />

RBA.<br />

Fibers are first modeled as jointed beams following the<br />

fluid mechanics in the forming section. The sheet<br />

<strong>structure</strong> obtained from this is representative <strong>of</strong> the wet<br />

sheet from the couch. The pressing simulation discretizes<br />

fibers into a number <strong>of</strong> solid elements around the lumen.<br />

Bonding between fibers is simulated using spring<br />

elements.<br />

The resulting fiber <strong>network</strong> was analyzed to determine<br />

its elastic modulus and deformation under small strains.<br />

The influence <strong>of</strong> fiber dimensions, namely fiber lengths,<br />

widths and thicknesses as well as bond stiffnesses on the<br />

elasticity <strong>of</strong> the <strong>network</strong> are studied. A brief account <strong>of</strong><br />

inclusion <strong>of</strong> fines, represented by individual cubical<br />

elements is also shown.<br />

ADDRESSES OF THE AUTHORS:<br />

S. Lavrykov (lavrykov@esf.edu) and B. V. Ramarao<br />

(bvramara@esf.edu): Empire State Paper Research<br />

Institute, State University <strong>of</strong> New York ESF, 1 Forestry<br />

Dr., Syracuse NY, USA 13210.<br />

S. B. Lindström (stefan.lindstroem@gmail.<strong>com</strong>):<br />

Mechanics, IEI, Institute <strong>of</strong> Technology, Linköping<br />

University, 583 81 Linköping, Sweden.<br />

K. M. Singh (kapil.singh@i<strong>paper</strong>.<strong>com</strong>): International<br />

Paper Co., Corporate Technology Center, 6283 Tri-Ridge<br />

Blvd., Loveland, OH 45140, USA.<br />

Corresponding author: B. V. Ramarao<br />

In order to investigate the impact <strong>of</strong> varying fiber<br />

dimensions or other similar properties on <strong>paper</strong><br />

properties, a number <strong>of</strong> <strong>com</strong>puter based <strong>simulations</strong> <strong>of</strong><br />

<strong>paper</strong> forming have been developed in the past. The<br />

earliest <strong>com</strong>puter <strong>simulations</strong> were <strong>of</strong> two dimensional<br />

<strong>paper</strong> <strong>structure</strong>s based on dropping straight or curvilinear<br />

random lines on a flat plane and investigating the<br />

statistics <strong>of</strong> the resulting <strong>structure</strong> (Kallmes, Corte 1960;<br />

Corte, Kallmes 1962). The next steps <strong>of</strong> 2D fiber<br />

<strong>network</strong>s were calculations <strong>of</strong> <strong>network</strong> mechanical<br />

properties. Reviews <strong>of</strong> these are found in Deng and<br />

Dodson, 1994 and by Bronkhorst, 2003. These 2D<br />

<strong>network</strong>s <strong>of</strong> the <strong>paper</strong> <strong>structure</strong> are usually obtained by<br />

placing the centers <strong>of</strong> fibers either uniformly randomly or<br />

with different flocculation rules on a rectangular or<br />

square area. Fiber orientation is adjusted by sampling<br />

from known distributions. For mechanical applications,<br />

256 Nordic Pulp and Paper Research Journal Vol 27 no.2/2012<br />

fibers were considered as beams - usually as simple<br />

Bernoulli beams (Heyden 2000; Heyden, Gustafsson<br />

2002) long free span, and using the Timoshenko model<br />

for short beams (Räisänen et al., 1996; Bronkhorst,<br />

2003). The fiber material is usually considered as elastic<br />

(Van den Akker, 1962; Deng, Dodson, 1994; Kahkonen,<br />

2003; Heyden 2000; Heyden, Gustafsson 2002). Some<br />

cases <strong>of</strong> elastic-plastic <strong>simulations</strong> are described in<br />

(Ramasubramanian, Perkins, 1987; Räisänen et al., 1996;<br />

Bronkhorst, 2003). The intersections <strong>of</strong> fibers are<br />

considered as connecting bonds which are either rigid or<br />

flexible, having some finite transitional and torsional<br />

stiffness. All 2D fibers models are easy to build and<br />

analyze, but the following issues render their applicability<br />

<strong>of</strong> questionable value for <strong>paper</strong> materials. It is impossible<br />

to calculate the thickness (and thus the density) <strong>of</strong> the<br />

simulated fiber web. Since <strong>paper</strong> sheet properties are<br />

strongly dependent on density, this is perhaps the most<br />

serious limitation <strong>of</strong> two dimensional models to simulate<br />

actual <strong>paper</strong> properties. Also, since most <strong>simulations</strong><br />

assume rigid inter-fiber bonds, the predictions always<br />

seriously overestimate the sheet’s elastic moduli (Van<br />

den Akker, 1963). The number <strong>of</strong> fiber bonds is also<br />

grossly overestimated, and therefore even in the case <strong>of</strong><br />

flexible bonds, the elastic modulus predictions are much<br />

higher than real values. Finally, beam theories are not<br />

accurate for beams with the extremely short free spans<br />

such as those occurring in <strong>paper</strong>.<br />

Another option to generate real fiber <strong>structure</strong>s is <strong>3D</strong><br />

<strong>network</strong>ing. Methods for creating three-dimensional fiber<br />

<strong>network</strong> <strong>structure</strong>s have been proposed by a number <strong>of</strong><br />

authors as discussed in detail by Alava and Niskanen,<br />

2006. Some <strong>of</strong> these models can be used for generation <strong>of</strong><br />

<strong>network</strong>s similar to hand sheets (Nilsen et al., 1998;<br />

Provatas et al., 2000; Heyden 2000; Heyden, Gustafsson<br />

2002; Vincent et al., 2009; Vincent et al., 2010), where<br />

fiber deposition occurs without considering<br />

hydrodynamic effects. Straight or curvilinear fibers are<br />

randomly generated inside an arbitrarily chosen volume,<br />

(Heyden 2000; Heyden, Gustafsson 2002), or generated<br />

over rectangular base and then deposited independently<br />

<strong>of</strong> each other. During deposition the shape <strong>of</strong> the fibers<br />

can change. In the KCL-PAKKA model (Niskanen et al.,<br />

1997; Nilsen et al., 1998) fibers do not bend but a relative<br />

shift <strong>of</strong> adjacent fiber segments can be provided<br />

depending on the stiffness <strong>of</strong> the wet fiber material and<br />

applied external pressure. In another case (Vincent et al.,<br />

2009; Vincent et al., 2010), fiber bending was modelled<br />

according to some artificial bending criteria (two<br />

prescribed bending angles were introduced, and these<br />

angles were supposed to be dependent on the wood<br />

species). An important consideration for <strong>paper</strong> <strong>network</strong>s<br />

is that the free fiber length segments are usually small<br />

such that bending deformations tend to be limited.<br />

Generations <strong>of</strong> artificial machine-made <strong>paper</strong> sheets,<br />

where fiber deposition was modeled with the influence <strong>of</strong><br />

water movement, were considered in work <strong>of</strong> Miettinen


PAPER PHYSICS<br />

et al. 2007; Switzer et al. 2004; Lindström, Uesaka 2008.<br />

Pulp fibers were modeled as multilink chains <strong>of</strong> rigid<br />

cylindrical segments with circular cross-section,<br />

immersed into Newtonian liquid. Fiber chains can bend<br />

and twist due to joints between segments. Different cases<br />

<strong>of</strong> liquid motion were considered, including simple<br />

drainage under gravity and external pressure, (see<br />

Miettinen et al. 2007; Switzer et al. 2004), and <strong>com</strong>plex<br />

water movement in different sections <strong>of</strong> roll-blade<br />

formers, (Lindström, Uesaka 2008; Lindstrom et al.,<br />

2009). These models suffer some disadvantages such as<br />

the limitations <strong>of</strong> rigid segments which do not permit the<br />

same fiber bending, the output <strong>network</strong>s have low density<br />

and higher thickness. However, these methods can be<br />

used for generating pulp webs <strong>of</strong> low consistency, and<br />

thus can simulate forming situations corresponding to the<br />

wet web before the wet pressing section in the <strong>paper</strong><br />

machine.<br />

The main goal <strong>of</strong> the present work is to develop a<br />

numerical approach to build a <strong>3D</strong> artificial hand sheet<br />

and machine sheet fiber <strong>network</strong>s with prescribed<br />

thickness and apparent density, consisting <strong>of</strong> collapsed<br />

and non-collapsed fibers, fines and fillers. We can take a<br />

simulated <strong>structure</strong> made by using multilink chains <strong>of</strong><br />

rigid cylindrical segments as fibers and abstract the <strong>3D</strong><br />

<strong>structure</strong> <strong>of</strong> the formed <strong>paper</strong> sheet. The abstraction is by<br />

using the center lines <strong>of</strong> the fibers and redecorating their<br />

thicknesses and widths to develop a fiber model with<br />

prismatic elements <strong>com</strong>prising the fibers. The fibers are<br />

also considered to be hollow to allow for the lumens and<br />

can be <strong>com</strong>pressed in a wet pressing simulation to<br />

generate a ‘final’ <strong>structure</strong>. In the following, we will<br />

describe our method <strong>of</strong> generating the <strong>paper</strong> <strong>structure</strong> and<br />

show how it is applied to a random ‘handsheet’ <strong>structure</strong><br />

and a realistic, simulated sheet formed using a twin wire<br />

roll former with blades. The <strong>structure</strong>s are then analyzed<br />

for their mechanical properties (in this case, the elastic<br />

modulus).<br />

Fiber <strong>network</strong> Generation<br />

Hand Sheet Formation<br />

A set <strong>of</strong> objects is defined for the simulation, which<br />

includes objects <strong>of</strong> three types: fibers, fines or fillers.<br />

These are characterized by their geometrical parameters<br />

which include length, width and thickness, their<br />

distributions and type <strong>of</strong> cross-section i.e. collapsed or<br />

non-collapsed, as in Fig 1. Fiber type 1a has only one<br />

finite element in thickness direction. This type was<br />

usually used to describe fines and fillers. Fiber types 1b<br />

and 1c have arbitrary number <strong>of</strong> elements in all three<br />

directions (length, width, thickness). Non-collapsed fiber<br />

type 1d has always one element in the thickness <strong>of</strong> fiber<br />

wall and arbitrary number <strong>of</strong> elements in other two<br />

directions. It is possible to include curl and kink <strong>of</strong> fibers<br />

by suitable modifications <strong>of</strong> these elementary <strong>structure</strong>s,<br />

although this was not implemented in the current<br />

simulation.<br />

The first step <strong>of</strong> the simulation is to generate an initial<br />

<strong>structure</strong>. Handsheet formation can be simulated by<br />

simply placing fibers (or objects) randomly in space<br />

(a)<br />

(c)<br />

(b)<br />

(d)<br />

Fig 1. Configurations <strong>of</strong> fibers used in the model<br />

Frequency<br />

0,45<br />

0,40<br />

0,35<br />

0,30<br />

0,25<br />

0,20<br />

0,15<br />

0,10<br />

0,05<br />

0,00<br />

Hardwood Pulp<br />

0,5 0,9 1,4 2,0 2,7<br />

Fiber length, mm<br />

Fig 2. Examples <strong>of</strong> fiber length distributions for s<strong>of</strong>twood and<br />

hardwood pulps<br />

according to their distribution in the furnish. An example<br />

<strong>of</strong> such distributions for hard wood and s<strong>of</strong>t wood pulps<br />

are shown in Fig 2.<br />

The second step is location <strong>of</strong> centers <strong>of</strong> objects in Z<br />

direction. Each new object has Z coordinate taking into<br />

account all objects already generated and located below<br />

this particular one. All points <strong>of</strong> such line have the same<br />

Z-coordinate.<br />

The third step is the generation <strong>of</strong> finite element grid in<br />

each object according to type <strong>of</strong> object cross-section and<br />

its dimensions. The topology <strong>of</strong> finite elements (shape<br />

and local numbering <strong>of</strong> nodes) should correspond to finite<br />

element s<strong>of</strong>tware used in the following steps. In our case<br />

this was the 8-node hexahedron solid element in the LS-<br />

DYNA Explicit program.<br />

Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 257


PAPER PHYSICS<br />

Fig 4. ZD view <strong>of</strong> simulated sample <strong>structure</strong> after forming in<br />

<strong>paper</strong> machine<br />

z max<br />

z<br />

z caliper<br />

Fig 3. An example <strong>of</strong> initial configuration <strong>of</strong> fiber <strong>network</strong> (all<br />

four fiber configurations and filler particles are presented)<br />

The fourth step is trimming out the part <strong>of</strong> finite element<br />

grid outside prescribed sample dimensions. A sample <strong>of</strong><br />

the finite element <strong>network</strong> generated after the fourth step<br />

is shown in Fig 3.<br />

And the fifth, the last step, is the calculation <strong>of</strong> fiber<br />

deposition velocities. All points <strong>of</strong> each particular object<br />

have the same initial velocity in vertical direction. The<br />

velocity is calculated from the deposition height and<br />

deposition time.<br />

Machine Sheet Formation<br />

This method <strong>of</strong> fiber <strong>network</strong> generation is described in<br />

detail by Lindström, Uesaka, 2008. Fibers are represented<br />

as sets <strong>of</strong> rigid segments connected by joints with bend<br />

and torsion stiffness. Mass and moment <strong>of</strong> inertia <strong>of</strong><br />

segments correspond to solid circular cylinders. Initially<br />

all fibers are generated randomly inside a prescribed<br />

volume, which is filled by a viscous in<strong>com</strong>pressible fluid.<br />

The motion <strong>of</strong> the fluid is governed by the threedimensional<br />

Navier–Stokes equations. Equations <strong>of</strong><br />

motion <strong>of</strong> fiber segments includes inertia forces,<br />

hydrodynamic drag forces from moving liquid and fiberto-fiber<br />

contact forces. The slurry is running through<br />

different sections <strong>of</strong> <strong>paper</strong> machine, where different<br />

boundary conditions are applied (Lindström, Uesaka,<br />

2008; Lindstrom et al., 2009). The final low consistency<br />

pulp mat is used as initial input data for the <strong>network</strong><br />

<strong>com</strong>pression model. Each circular fiber is replaced by the<br />

collapsed or non-collapsed fiber, Fig 1, and is divided<br />

into a set <strong>of</strong> finite elements. A mechanical material model<br />

(elastic, elastic-plastic, visco-elastic-plastic) is assigned<br />

for each particular fiber. The wet fiber flexibility is a<br />

fiber parameter that has been investigated extensively in<br />

the past and can be used to determine a representative<br />

fiber modulus. Since wet fibers are substantially plastic, it<br />

is necessary to define plasticity parameters in addition to<br />

the elasticity. Fibers are anisotropic and their deformation<br />

is considerably <strong>com</strong>plex. However, for an initial<br />

simulation such as this, we used a simple isotropic<br />

bilinear elastic-plastic model with the modulus<br />

determined from the wet fiber flexibility data in the<br />

literature [Paavilainen, Luner, 1986; Abitz et al., 1985].<br />

An example <strong>of</strong> machine-generated <strong>network</strong> is shown in<br />

Fig 4.<br />

258 Nordic Pulp and Paper Research Journal Vol 27 no.2/2012<br />

t load t rest t unload<br />

Fig 5. Upper rigid plate movement history during mat<br />

<strong>com</strong>pression<br />

Fiber <strong>network</strong> Compression<br />

The simulation <strong>of</strong> fiber mat <strong>com</strong>pression was performed<br />

with the explicit finite element method, [see Zhong 1993<br />

and Hallquist 2005]. We used LS-DYNA, a program for<br />

Nonlinear Dynamic Analysis <strong>of</strong> Structures in Three<br />

Dimensions, Version 971 (Hallquist 2005). The 8 node<br />

hexahedron elements with three displacements in each<br />

node were used to represent fiber <strong>network</strong>. The central<br />

finite difference scheme was used for the numerical<br />

integration <strong>of</strong> the equations <strong>of</strong> motion. Fiber-to-fiber<br />

interactions during simulation were calculated from<br />

balance <strong>of</strong> nodal inertia forces, external forces, internal<br />

forces and contact forces. All contact couples were found<br />

automatically during simulation. (This method <strong>of</strong> contact<br />

handling was made possible by using a keyword<br />

*CONTACT_AUTOMATIC_GENERAL).<br />

To obtain the fiber mat with prescribed thickness and<br />

density, the <strong>network</strong> was <strong>com</strong>pressed between two rigid<br />

plates. The non-moving plate is located below the<br />

previously generated fiber <strong>network</strong>. The moving plate<br />

with prescribed motion in vertical direction (Fig 5) is<br />

initially located on the top <strong>of</strong> the mat. Fiber settling and<br />

<strong>com</strong>pression are performed during a loading time ,<br />

which was equal to 1-2 ms. A stabilization time is<br />

need to decrease dynamic effects in the <strong>com</strong>pressed fiber<br />

mat which was generally set to 1 ms. An unloading time<br />

is also needed to remove the upper plate and to<br />

give the fiber mat possibility to expand in thickness<br />

direction due to internal stresses in fibers. The unloading<br />

time was 0.1-1.0 ms usually.<br />

The results <strong>of</strong> some <strong>3D</strong> <strong>network</strong> <strong>simulations</strong> are<br />

presented in Fig 6-9. The initial fibers configuration from<br />

Fig 3 after <strong>com</strong>pression is shown in Fig 6. A larger scale<br />

model with about 1500 fiber <strong>of</strong> the mixture <strong>of</strong> hardwood<br />

and s<strong>of</strong>twood pulps is presented in Fig 7. Results <strong>of</strong><br />

simulation <strong>of</strong> fiber <strong>network</strong> with high content <strong>of</strong> mineral<br />

filler (25% weight fraction) are shown in Fig 8 and 9. A<br />

layer created by fillers is observed in the pictures. The<br />

example <strong>of</strong> machine-made fiber <strong>network</strong> after<br />

<strong>com</strong>pression is shown in Fig 10. The colors serve to<br />

differentiate the fibers in the <strong>network</strong>.<br />

t


PAPER PHYSICS<br />

Fig 10. Simulated machine-made sample after <strong>com</strong>pression<br />

Fig 6. Fiber <strong>network</strong> from Fig. 3 after <strong>com</strong>pression<br />

ΔX<br />

ΔX<br />

L<br />

Fig 7. The sample with 50% mixture <strong>of</strong> s<strong>of</strong>twood and hardwood<br />

pulps<br />

Fig 8. Fiber <strong>network</strong> with high filler content – an example <strong>of</strong><br />

coating simulation<br />

Fig 9. Filler particles distribution in the fiber mat during coating<br />

formation<br />

Fig 11. Tensile test for elastic modulus calculations<br />

Fiber Network Applications<br />

Calculation <strong>of</strong> Apparent Elastic Modulus<br />

The apparent elastic modulus is calculated the same way<br />

as usually calculated from the physical tensile test. The<br />

sample length should be at least 1.2 times bigger than the<br />

longest fiber in the <strong>network</strong>, as re<strong>com</strong>mended in Heyden<br />

2000. The example <strong>of</strong> such sample is presented in Fig 11<br />

(sample length is 5 mm, width is 1 mm). Material <strong>of</strong><br />

fibers is isotropic elastic. Mechanical properties <strong>of</strong> single<br />

fibers <strong>of</strong> different dry wooden species can be found in<br />

(Bronkhorst 2003; Page et al. 1977; Katz et al 2008).<br />

Before the solution <strong>of</strong> elastic boundary-value problem,<br />

the <strong>network</strong> should be analyzed and bonds between fibers<br />

should be set up. A contact couple can be defined only<br />

between two elements which belong to two different<br />

fibers. From one to four links (bonds) may be set up in<br />

one contact couple. The stiffness <strong>of</strong> each bond is<br />

calculated from the value <strong>of</strong> contact area (overlapped area<br />

<strong>of</strong> two contact surfaces) and bonding stiffness,<br />

experimentally found by Thorpe et al. 1976. The average<br />

maximum shear stress reported by Thorpe et al. 1976 and<br />

Perkins 2001, for the holocellulose bond was<br />

corresponding to an average maximum strain<br />

. The bond stiffness parameter is obtained<br />

as<br />

and is equal to the<br />

spring force between each pair <strong>of</strong> bonding nodes. Fibers<br />

not bonded to other fibers <strong>of</strong> the <strong>network</strong> are eliminated<br />

from consideration. To solve the elastic problem, the<br />

kinematic boundary conditions should be applied to the<br />

finite element grid. Prescribed displacements are assigned<br />

to the nodes located on clamped area on the top and the<br />

bottom surfaces from two opposite sides <strong>of</strong> the sample,<br />

Fig 11.<br />

After solution <strong>of</strong> elastic problem, nodal displacements<br />

are used for calculation <strong>of</strong> strain and stress distributions<br />

in finite elements.<br />

Integration <strong>of</strong> stresses over element’s area in some<br />

sample cross-sections gives average force and average<br />

Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 259


PAPER PHYSICS<br />

Elastic Modulus, GPa<br />

7,0<br />

6,0<br />

5,0<br />

4,0<br />

3,0<br />

2,0<br />

1,0<br />

S<strong>of</strong>twood<br />

Hardwood<br />

0,0<br />

0,09 0,11 0,13 0,15<br />

Caliper, mm<br />

Fig 12. Elastic modulus <strong>of</strong> <strong>network</strong>s <strong>of</strong> s<strong>of</strong>twood and hardwood<br />

pulps<br />

Elastic Modulus, GPa<br />

4,0<br />

3,0<br />

2,0<br />

1,0<br />

0,0<br />

5,0 7,5 10,0 12,5 15,0<br />

Fiber Thickness, µm<br />

Fig 13. Elastic modulus <strong>of</strong> simulated <strong>network</strong> as a function <strong>of</strong><br />

fiber thickness<br />

Elastic Modulus, GPa<br />

4,0<br />

3,0<br />

2,0<br />

1,0<br />

0,0<br />

10,0 15,0 20,0 25,0 30,0 35,0 40,0<br />

Fiber Width, µm<br />

Fig 14. Elastic modulus <strong>of</strong> simulated <strong>network</strong> as a function <strong>of</strong><br />

fiber width<br />

Elastic Modulus, GPa<br />

6,0<br />

5,0<br />

4,0<br />

3,0<br />

2,0<br />

1,0<br />

0,0<br />

0,0 0,3 0,5 0,8 1,0 1,3 1,5 1,8<br />

Fiber Length, mm<br />

Fig 15. Influence <strong>of</strong> bond stiffness on elastic modulus <strong>of</strong> fiber<br />

<strong>network</strong>s<br />

0.01<br />

0.05<br />

0.1<br />

0.2<br />

0.3<br />

0.4<br />

0.5<br />

Elastic Modulus, GPa<br />

6,00<br />

5,95<br />

5,90<br />

5,85<br />

5,80<br />

0,5 1,0 1,5 2,0 2,5<br />

Sample Length/Max Fiber Length<br />

Fig 16. Influence <strong>of</strong> sample length on elastic modulus <strong>of</strong> 60 gsm<br />

fiber <strong>network</strong> (fiber length 2 mm, sheet caliper 0.1 mm)<br />

stress in the sample. The strain value is equal to ratio <strong>of</strong><br />

prescribed displacement to the length <strong>of</strong> free span<br />

(distance between clamps), so the apparent elastic<br />

modulus is found. Fig 12 shows the elastic modulus for<br />

sample <strong>network</strong>s <strong>of</strong> s<strong>of</strong>twood and hardwood pulps with<br />

length distributions shown in Fig 2. The modulus is<br />

shown as a function <strong>of</strong> sheet caliper corresponding to<br />

different wet pressing levels. We observe that the<br />

modulus increases as the caliper is decreased, i.e. the<br />

sheet is densified by wet pressing to different levels. It<br />

can also be seen that the s<strong>of</strong>twood fiber <strong>network</strong> shows<br />

higher modulus as <strong>com</strong>pared to the hardwood fibers. This<br />

result clarifies the impact <strong>of</strong> fiber length on sheet<br />

modulus. In order to evaluate the influence <strong>of</strong> some fiber<br />

parameters including fiber width, fiber thickness and<br />

bonding stiffness, on apparent tensile elastic modulus <strong>of</strong><br />

fiber mat we conducted more <strong>simulations</strong> for fiber mats<br />

with the same basis weight 60 gsm and the same material<br />

properties. The results are shown in Figs 12-15 and in<br />

Table 1.<br />

Fig 13 and Fig 14 show that the elastic modulus<br />

decreases with increased thickness (evaluated at constant<br />

caliper). Increasing thickness results in lesser fibers in the<br />

sheets, resulting in the lower values observed. Fig 15<br />

shows the influence <strong>of</strong> the bond stiffness; increased<br />

stiffness shows increases in the elastic modulus.<br />

However, the increase quickly saturates and when the<br />

bond stiffness increases beyond 0.3, gains in the modulus<br />

are small.<br />

The size <strong>of</strong> the <strong>network</strong> for simulation <strong>of</strong> the elastic<br />

properties should be sufficiently large such that the<br />

effects <strong>of</strong> singular fibers spanning a large portion <strong>of</strong> the<br />

<strong>network</strong> are minimized. In order to investigate this<br />

effect, we carried out <strong>simulations</strong> <strong>of</strong> the elastic modulus<br />

with increasing <strong>network</strong> length (denoted as sample size)<br />

keeping fiber length a constant. Fig 16 shows the results<br />

for fibers <strong>of</strong> length 2mm. The simulation result for the<br />

elastic modulus is independent <strong>of</strong> the <strong>network</strong> size for<br />

<strong>network</strong>s greater than 1.5 fiber lengths. Fig 17 shows a<br />

graph <strong>of</strong> the wet pressing pressure as a function <strong>of</strong> mat<br />

density for an example s<strong>of</strong>twood and hardwood fibers.<br />

These are representative calculations indicating that for a<br />

given wet pressing pressure, the s<strong>of</strong>twood sheets are<br />

more densified than the hardwoods. The calculations<br />

serve as illustrative cases since the parameters chosen for<br />

the fibers were arbitrary.<br />

260 Nordic Pulp and Paper Research Journal Vol 27 no.2/2012


PAPER PHYSICS<br />

Table 1. Properties <strong>of</strong> fibers and simulated 60 gsm sheets, 1x2 mm sample (20F represents fines, fiber objects consisting <strong>of</strong> single<br />

elements).<br />

Parameter<br />

S<strong>of</strong>twood pulp Mix 1<br />

Mix 2<br />

Mix 3 Hardwood pulp<br />

0HW+100SW 50HW+50SW 64HW+16SW+20F 80HW+20SW 100HW+0SW<br />

Fiber Length (Avg.), mm 3.0 2.0 1.2 1.4 1.0<br />

Fiber Width (Avg.), μm 24 24 24 24 24<br />

Elastic Modulus Fiber, GPa 10 10 10 10 10<br />

Caliper, μm 100 100 100 100 100<br />

Number <strong>of</strong> fibers 612 783 1162 952 1111<br />

Elastic Modulus <strong>of</strong> sheet, GPa 6.433 5.920 4.309 4.278 2.959<br />

Stiffness (mN·m), simulated 0.536 0.493 0.359 0.356 0.246<br />

Stiffness (mN·m), experiment<br />

(Lavrykov et al., 2010)<br />

0.763 0.703 0.693<br />

Wet Pressing Pressure, MPa<br />

60,0<br />

50,0<br />

40,0<br />

30,0<br />

20,0<br />

10,0<br />

S<strong>of</strong>twood<br />

Hardwood<br />

0,0<br />

0,4 0,5 0,6 0,7 0,8<br />

Density, g/cm 3<br />

Fig 17. Pressure as a function <strong>of</strong> mat density<br />

Table 1 below shows the results <strong>of</strong> <strong>simulations</strong> <strong>of</strong> 60<br />

gsm sheets <strong>com</strong>posed <strong>of</strong> a mix <strong>of</strong> hardwood and<br />

s<strong>of</strong>twood pulps in different ratios along with a portion <strong>of</strong><br />

the fines. Note that the mixes are shown in number<br />

fractions (not mass) as is usual for experimental data. The<br />

simulation results show that the sheet modulus is a strong<br />

function <strong>of</strong> the s<strong>of</strong>twood content, increasing in magnitude<br />

as this increases. An important parameter, the sheet<br />

stiffness was also evaluated and shown to be a strong<br />

function <strong>of</strong> the s<strong>of</strong>twood content in the pulp. The final<br />

rows in this table present stiffness estimates using the<br />

<strong>simulations</strong> as <strong>com</strong>pared to data obtained from<br />

experiments. It appears that when the elastic modulus <strong>of</strong><br />

the fibers is assumed to be the same, the longer fibers<br />

yield a higher stiffness value, a trend that corresponded to<br />

the experimental results. The magnitudes <strong>of</strong> the<br />

stiffnesses observed experimentally are much higher than<br />

the simulation predictions though. The modulus <strong>of</strong> the<br />

fibers and other parameters were obtained from published<br />

values in the literature rather than measurements or<br />

estimates on the furnish itself. This would account for the<br />

difference in magnitude <strong>of</strong> the estimates.<br />

Simulation <strong>of</strong> Wet Pressing <strong>of</strong> Sheets<br />

The most significant use <strong>of</strong> simulation is to predict the<br />

<strong>structure</strong> <strong>of</strong> the <strong>paper</strong> sheets with particular reference to<br />

their z-dimensions i.e. the caliper or analogously, their<br />

density. This is not possible with 2D <strong>simulations</strong> and also<br />

simplistic constructions allowing the fibers to bend<br />

according to external rules. The best method <strong>of</strong><br />

simulation is to determine the fiber reaction force during<br />

<strong>com</strong>pression, the so-called <strong>com</strong>pressive stress and also<br />

simultaneously track the drag force exerted by the<br />

moving fluid. This is possible with a particle level<br />

simulation such as the present one. The fiber<br />

displacements were simulated using LS-DYNA. At each<br />

step <strong>of</strong> pressing, the fiber reaction force was summed and<br />

used as the total <strong>com</strong>pressive stress borne by the fibrous<br />

<strong>structure</strong>. The <strong>com</strong>bination <strong>of</strong> the drag, translated into<br />

permeability and the <strong>com</strong>pressive stress was applied to a<br />

homogenized wet pressing model (Lavrykov et al., 2009)<br />

to determine the sheet caliper at different values <strong>of</strong><br />

applied pressure. The hydraulic stress was estimated<br />

using an effective medium type approximation with<br />

periodic cells, adjusted for local change in porosity.<br />

These were <strong>com</strong>bined to generate the <strong>structure</strong> in a timeexplicit<br />

method. Fillers (or equivalent fines) were<br />

considered as single discrete elements <strong>of</strong> approximately 2<br />

to 4µm in size. This allowed the calculation <strong>of</strong> the sheet<br />

caliper and density at given levels <strong>of</strong> applied pressure and<br />

the as a result, the sheet <strong>structure</strong>.<br />

Discussion<br />

As was mentioned before, the main objective <strong>of</strong> this work<br />

was to provide a new method for the generation <strong>of</strong> fiber<br />

<strong>network</strong>s which can be applicable to numerical solution<br />

<strong>of</strong> mechanical problems. This means fibers in an artificial<br />

mat should consist <strong>of</strong> set <strong>of</strong> finite elements connected in<br />

nodes. During mat formation fibers should not be bent<br />

only but <strong>com</strong>pressed too taking into account possible<br />

fiber collapsing. In this situation all previously reported<br />

methods (Nilsen et al., 1998; Provatas et al., 2000;<br />

Heyden 2000; Heyden, Gustafsson 2002; Vincent et al.,<br />

2009; Vincent et al., 2010) were found not correspondent<br />

to our goals.<br />

To provide a simulation <strong>of</strong> fibers settling and<br />

<strong>com</strong>pression the explicit analysis s<strong>of</strong>tware was used.<br />

There are several important reasons for using explicit<br />

analysis instead <strong>of</strong> the implicit one for solution <strong>of</strong> this<br />

type <strong>of</strong> problems. First, the explicit methods perform this<br />

analysis much faster. For example, for the sample in Fig<br />

7, which contains about 2000 fibers with 120,000<br />

elements and 500,000 nodes, the solution <strong>of</strong> the static<br />

elastic problem using the implicit method on a PC<br />

<strong>com</strong>puter with two Xeon Quad Core 3 GHz processors<br />

and 24 GB RAM under 64-bit Windows 7 operational<br />

system requires 4 Hrs. During solution <strong>of</strong> nonlinear<br />

elastic-plastic contact problem with large strains and<br />

Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 261


PAPER PHYSICS<br />

displacements similar time is required for each iteration<br />

at each loading step. At the same time, the total<br />

<strong>com</strong>putational time for explicit analysis <strong>of</strong> this particular<br />

model was only approximately 6 hrs. Of course, the total<br />

<strong>com</strong>putational time for explicit analysis is highly<br />

dependent <strong>of</strong> integration time step which, in turn, is<br />

dependent <strong>of</strong> element size and mechanical properties <strong>of</strong><br />

fiber material - wet fiber density, wet fiber Young’s<br />

elastic modulus and Poisson’s ratio, (Hallquist 2005).<br />

Using <strong>of</strong> mass scaling procedure in LS-DYNA (Hallquist<br />

2005) can significantly reduce <strong>com</strong>putational time, but<br />

we did not use this option.<br />

Second, during solution <strong>of</strong> contact problems with<br />

implicit methods using <strong>com</strong>mercial finite element<br />

s<strong>of</strong>tware systems like Ansys or Abaqus it is necessary to<br />

define all contact couples in advance. In fiber <strong>network</strong><br />

<strong>com</strong>pression <strong>simulations</strong>, the fiber movement is<br />

unpredictable so contact couples are unknown. In the<br />

same time, an automatic contact search option is exist in<br />

explicit analysis s<strong>of</strong>tware LS-DYNA.<br />

The simulation itself provides a good tool to understand<br />

the impact <strong>of</strong> furnish <strong>com</strong>position and its changes on<br />

sheet properties. In addition, by matching the technique<br />

with <strong>simulations</strong> <strong>of</strong> the forming section, more <strong>com</strong>plete<br />

<strong>simulations</strong> can be constructed which reflect machine<br />

operating parameters in addition to the stock variables.<br />

Conclusions<br />

A <strong>com</strong>prehensive <strong>network</strong> simulation <strong>of</strong> sheet forming<br />

was conducted. The <strong>simulations</strong> show the significance <strong>of</strong><br />

using different fiber level descriptions <strong>of</strong> the forming<br />

process in the wet end and wet pressing to construct the<br />

<strong>structure</strong>. The resulting <strong>structure</strong>s can be used for<br />

simulating a variety <strong>of</strong> transport and mechanical<br />

properties <strong>of</strong> <strong>paper</strong> sheets, and in particular, the effects <strong>of</strong><br />

different fiber furnishes and their treatments. This enables<br />

the optimization <strong>of</strong> furnish and processing to obtain high<br />

performance <strong>structure</strong>s.<br />

Acknowledgements<br />

We would like to thank the member <strong>com</strong>panies <strong>of</strong> the<br />

Empire State Paper Research Associates, SUNY ESF,<br />

Syracuse NY for partial funding <strong>of</strong> this work.<br />

Literature<br />

Abitz, P. R., Cresson, T., Brown, A. F. and Luner, P. (1987):<br />

The Role <strong>of</strong> Wet-Fiber Flexibility in Governing Wet Web<br />

Properties – A Preliminary Report, Empire State Paper<br />

Research Associates, Research Report No. 85, Chapter 6,<br />

SUNY College <strong>of</strong> Env. Sci. Forestry, Syracuse NY, USA, 97-<br />

110.<br />

Alava, M. and Niskanen, K. (2006): The Physics <strong>of</strong> Paper,<br />

Rep. Prog. Phys., 69(3), 669-723.<br />

Bronkhorst, C.A. (2003): Modeling Paper as a Twodimensional<br />

Elastic-plastic Stochastic Network, Int. J. Solids<br />

Struct., 40(20), 5441-5454.<br />

Corte, H. and Kallmes, O.J. (1962): Statistical Geometry <strong>of</strong> a<br />

Fibrous Network, In: Formation and Structure <strong>of</strong> Paper, Trans.<br />

2 nd Fund. Res. Symp. Oxford 1961, 13-46.<br />

Deng, M. and Dodson, C.T.J. (1994): Paper: an Engineered<br />

Stochastic Structure, TAPPI Press.<br />

Hallquist J.O. (2005): LS-DYNA Theory Manual, Livermore<br />

S<strong>of</strong>tware Technology Corp.<br />

Heyden S. (2000): Network Modeling for the Evaluation <strong>of</strong><br />

Mechanical Properties <strong>of</strong> Cellulose Fibre Fluff, PhD thesis,<br />

Lund University, Lund, Sweden.<br />

Heyden, S. and Gustafsson, P.J. (2002): Stress-strain<br />

Performance <strong>of</strong> Paper and Fluff by Network Modelling, In: The<br />

Science <strong>of</strong> Papermaking, Trans. 12 th Fund. Res. Symp. Oxford<br />

2001, Bury, UK, PPFRS, 1385-1401.<br />

Kahkonen, S. (2003): Elasticity and Stiffness Evolution in<br />

Random Fibre Networks, PhL thesis, University <strong>of</strong> Jyvaskyla.<br />

Kallmes, O. and Corte, H. (1960): The Structure <strong>of</strong> Paper. I.<br />

The Statistical Geometry <strong>of</strong> an Ideal Two-dimensional Fibre<br />

Network, Tappi, 43(9), 737-752.<br />

Katz, J. L., Spencer P., Wang Y., Misra A., Marangos O. and<br />

Friis L. (2008): On the Anisotropic Elastic Properties <strong>of</strong> Woods,<br />

J. Mater. Sci., 43(1), 139-145.<br />

Lavrykov, S., Mishra, G.K. and Ramarao, B. V., P. (2010):<br />

The Stiffness <strong>of</strong> Paper – Fibers vs. Bonds, Empire State Paper<br />

Research Associates, Research Report No. 132, Chapter 3,<br />

SUNY College <strong>of</strong> Env. Sci. Forestry, Syracuse NY, USA, 13-22.<br />

Lavrykov, S., Singh, R. and Ramarao, B. V. (2009):<br />

Compressiblity effects in the consolidation <strong>of</strong> pulp mats with<br />

filler particles, In: Proceedings <strong>of</strong> 14 th Fundamental Research<br />

Symposium, Fundamental Research Society, Bury Lancashire,<br />

UK.<br />

Lindström, S.B. and Uesaka, T. (2008): Particle-level<br />

Simulation <strong>of</strong> Forming <strong>of</strong> the Fiber Network in Papermaking, Int.<br />

J. Eng. Sci., 46(9), 858-876.<br />

Lindström, S. B., Uesaka, T. and Hirn, U. (2009): Evolution <strong>of</strong><br />

the <strong>paper</strong> <strong>structure</strong> along the length <strong>of</strong> a twin-wire former, In:<br />

Advances in Pulp and Paper Research, Trans. 14 th Fund. Res.<br />

Symp., Oxford 2009, Bury, UK, PPFRS, 207-245.<br />

Miettinen, P.P.J., Ketoja, J.A. and Klingenberg, D.J. (2007):<br />

Simulated Strength <strong>of</strong> Wet Fibre Networks, J. Pulp Paper Sci.,<br />

33(4), 198-205.<br />

Niskanen, K., Nilsen, N., Hellen, E. and Alava, M. (1997):<br />

KCL-PAKKA: Simulation <strong>of</strong> the <strong>3D</strong> Structure <strong>of</strong> Paper, In: The<br />

Fundamentals <strong>of</strong> Papermaking Materials, Trans. 11 th Fund.<br />

Res. Symp. Cambridge 1997, Leatherhead Surrey, PIRA<br />

International, 1273-1291.<br />

Nilsen, N., Zabihian, M. and Niskanen, K. (1998): KCL-<br />

PAKKA: a Tool for Simulating Paper Properties, Tappi J., 81<br />

(5), 163-166.<br />

Page, D.H., El-Hosseiny, F. and Lancaster, A.P.S. (1977):<br />

Elastic Modulus <strong>of</strong> Single Wood Pulp Fibers, Tappi, 60(4), 114-<br />

117.<br />

Paavilainen, L. and Luner, P. (1986): Wet-Fiber Flexibility as<br />

a Predictor <strong>of</strong> Sheet Properties, Empire State Paper Research<br />

Associates, Research Report No. 84, Chapter 9, SUNY College<br />

<strong>of</strong> Env. Sci. Forestry, Syracuse NY, USA, 151-172.<br />

262 Nordic Pulp and Paper Research Journal Vol 27 no.2/2012


PAPER PHYSICS<br />

Perkins, R.W. Jr. (2001): Models for Describing the Elastic,<br />

Viscoelastic, and Inelastic Mechanical Behavior <strong>of</strong> Paper and<br />

Board, In: Mark, R.E., Habeger, C.C., Borch, J., Lyne, M.B.<br />

(Eds.), Handbook <strong>of</strong> Physical Testing <strong>of</strong> Paper, vol. 1. Marcel<br />

Dekker, New York, 1-76.<br />

Ramasubramanian, M. K. and Perkins, R.W. Jr. (1987):<br />

Computer simulation <strong>of</strong> the Uniaxial Elastic-Plastic Behavior <strong>of</strong><br />

Paper, Empire State Paper Research Associates, Research<br />

Report No. 87, Chapter 4, 44-79. SUNY College <strong>of</strong> Env. Sci.<br />

Forestry, Syracuse NY, USA.<br />

Provatas, N., Haataja, M., Asikainen, J., Majaniemi, S.,<br />

Alava, M. and Ala-Nissila, T. (2000): Fiber Deposition Models<br />

in Two and Three Spatial Dimensions, Colloids and Surfaces,<br />

A: Physicochemical and Engineering Aspects, 165(1-3), 209-<br />

229.<br />

Räisänen, V.I., Alava, M.J., Nieminen, R.M. and Niskanen,<br />

K.J. (1996): Elastic-plastic Behaviour in Fibre Networks, Nordic<br />

Pulp and Paper Research Journal, 11(4), 243-248.<br />

Switzer III, L.H., Klingenberg, D.J. and Scott, C.T. (2004):<br />

Handsheet Formation and Mechanical Testing via Fiber-level<br />

Simulations”, Nordic Pulp and Paper Research Journal, 19(4),<br />

434-439.<br />

Thorpe, J.L., Mark, R.E., Eusufzai, A.R.K. and Perkins, R.W.<br />

(1976): Mechanical Properties <strong>of</strong> Fiber Bonds, Tappi, 59(5), 96-<br />

100.<br />

Van Den Akker, J.A. (1962): Some Theoretical Considerations<br />

on the Mechanical Properties <strong>of</strong> Fibrous Structures, In:<br />

Formation and Structure <strong>of</strong> Paper, Trans. 2 nd Fund. Res. Symp.<br />

Oxford, 1961, 205-241.<br />

Vincent, R., Rueff, M. and Voillot, C. (2009): 3-D<br />

Computational Simulation <strong>of</strong> Paper Handsheet Structure and<br />

Prediction <strong>of</strong> Apparent Density“, Tappi J., 8 (9), 10-17.<br />

Vincent, R., Rueff, M., and Voillot, C. (2010): Prediction <strong>of</strong><br />

Handsheet Tensile Strength by Computational Simulation <strong>of</strong><br />

Structure, Tappi J., 9 (1), 15-19.<br />

Zhong, Z.H. (1993): Finite Element Procedures for Contact-<br />

Impact Problems, Oxford Univ. Press.<br />

Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 263

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!