Professional Documents
Culture Documents
and
Philip E. Humphrey
Associate Professor
College of Forestry
Oregon State University
Corvallis, OR 97331-7402
(Received July 2002)
ABSTRACT
The quantitative and economical importance of continuous pressing has increased steadily over the
last two decades. An analytical model will be helpful in better understanding and improving this
process and in developing new manufacturing techniques. Reported here is a simulation model that
accounts for heat and mass transfer as well as rheological and adhesion mechanisms that occur in
three dimensions as mat material passes through a continuous press. The model enables one to predict
the evolution of important variables including temperature, moisture content, air and water vapor
pressure, density profile, and adhesive bond strength. The scientific principles upon which the model
depends are summarized along with the boundary conditions and modeling strategy employed. Model
predictions are presented and discussed for a typical medium density fiberboard (MDF) production
plant.
Keywords: Wood-based composites, continuous pressing, computer simulation, thermodynamics,
rheology, adhesion, density profile development, product and process innovations.
the foundation on which to develop analytical components will be summarized in turn; der-
process models based on fundamental princi- ivations and justifications of model assump-
ples. tions and mathematical formulations are not
A few analytical models of hot pressing included here; they are dealt with in Hum-
have been presented during the last two de- phrey and Bolton (1989a), Thoemen and
cades. Thoemen (2000) gives a comprehensive Humphrey (2003a, b), and Thoemen (2000).
literature review of such models. All models For modeling purposes, the wood-furnish
published to date simulate batch presses. Con- mat is considered as a homogeneous material.
tinuous pressing has, however, been increasing That is to say, a macroscopic view is used
steadily in quantitative and economical im- where the size of the representative volume is
portance. Not only does the productivity of large relative to that of the wood constituents.
this technology exceed that of batch pressing, However, a microscopic view is used when-
but it also has opened new opportunities to ever appropriate to illustrate model assump-
manipulate panel properties. tions or to explain specific features of the
Existing models for batch presses can be ex- modeled system.
trapolated to the continuous press to a certain
degree, the physical mechanisms within the Summary of mass and heat
mat being the same in both types of machine, transfer mechanisms
but there are serious limitations to so doing.
Not all of the important features of continuous The wood-furnish mat is a capillary-porous
pressing can be simulated by modifying con- material consisting of natural fiber cell wall
ventional models of the batch pressing pro- material and inter- and intra-particle voids,
cess. In the continuous press, for example, the which form a system of interlinked cavities.
boundary conditions change in the feed direc- The gas in the pore spaces is regarded as a
tion. Furthermore, the gas mixture in the con- pure two-component mixture of air and water
tinuous press escapes not only through the vapor. Measurements reported by Denisov et
sides of the mat, but also through its surfaces al. (1975) suggest that the proportion of other
immediately in front of and behind the press. components is relatively small; their influence
The model presented here can be applied to on the gas properties and mass and heat trans-
both batch and continuous pressing processes; fer rates is therefore unlikely to be significant.
it is summarized in this paper, with those as- Future versions of the model will, however,
pects that are specific to the continuous press account for the generation and movement of
presented in more detail. The interested reader substances such as volatile organic compounds
is referred to Thoemen and Humphrey (2003a, and formaldehyde, since these may be envi-
b) for a comprehensive description of the ba- ronmentally important and may play roles in
sics of the model and the numerical proce- the kinetics of adhesive cure or chemical mod-
dures employed. Predictions of interfiber bond ification of fibers.
strength development, based on adhesion ki- The model accounts for three-dimensional
netics data derived with the ABES technique convective and diffusive gas flow through the
(Humphrey 1999), are not included here; they void system of the mat. Bound water diffu-
will be dealt with in a separate publication. sion, surface diffusion, and the translation of
water condensed in small inter-particle capil-
laries are unlikely to be significant relative to
MODELING MASS AND HEAT TRANSFER
water vapor translation through the void sys-
AND RHEOLOGY
tem; they are not included, therefore, in the
The model consists of a three-dimensional model.
heat and mass transfer model, combined with Convective fluxes of the gas mixture devel-
a one-dimensional rheological model. Both op in response to total gas pressure gradients
WOOD AND FIBER SCIENCE
Monday Jun 09 2003 10:55 AM wood 35_403 Mp_458
Allen Press • DTPro System
File # 03em
and can be calculated by applying Darcy’s Both conductive and convective heat trans-
law. Total pressure gradients develop between fer in three dimensions are accounted for. Con-
the surfaces and the central plane of the mat ductive heat fluxes are proportional to tem-
and between its horizontal center and edges. perature gradients and can be described by
While the horizontal gradients may be consid- Fourier’s law:
erable, the cross-sectional ones usually dissi-
q 5 2kt¹T (4)
pate almost instantaneously in particulate
composites (Thoemen 2000). Darcy’s law is where:
used in the form:
q conductive heat flux [J m22 m21]
kp Mi kt thermal conductivity [W m21 K21]
jic 5 2 p ¹p i 5 v, a (1)
h RTabs i T temperature [8C or K]
where: Convective heat transfer is associated with
the gas flow (including diffusion) through the
v, a subscripts to denote vapor or air mat. In principle, two mechanisms contribute
jic convective flux of component i [kg to convective heat transfer. First, phase change
m22 s21] of water always accompanies the release or
Mi molecular mass of component i [kg consumption of latent heat of evaporation and
mol 21] latent heat of sorption. Second, gas that moves
pi partial pressure of component i [Pa] from a hot to a cold region releases sensible
p total gas pressure [Pa] heat, just as cold gas takes up heat energy
R gas constant [J mol 21 K21] when moving into a warmer region. The con-
Tabs absolute temperature [K] tribution of the transfer of sensible heat to
kp permeability coefficient [m 2] overall heat flow is small, however, because
h dynamic viscosity of the gas mixture the mass of vapor and air is very small com-
[Pa s] pared to the mass of wood and bound water.
In addition, diffusive fluxes occur that are The transfer of sensible heat is therefore not
caused by partial pressure gradients of the gas included.
components. This mechanism likely contrib- Two moisture phases are recognized: water
utes little to overall mass transfer during hot bound in the cell-wall material and water va-
pressing. Considerable partial pressure gradi- por in the voids; free water is not considered
ents of air and water vapor develop between in the present approach. The moisture is as-
the surfaces and the central plane of the mat sumed to be uniformly distributed within sin-
however, at least temporarily. Diffusive fluxes gle wood particles, and also in local equilib-
can be described by Fick’s law: rium with the vapor phase in the surrounding
voids. This assumption appears to be justified
Mi for small particles, such as single wood fibers.
jid 5 2Deff ¹p i 5 v, a (2)
RTabs i However, when moving to simulate materials
with large particle sizes (e.g., strands), it may
where:
become necessary to revise this assumption.
jid diffusive flux of component i [kg m22 The relationship between equilibrium moisture
s21] content (EMC), within-void relative humidity,
Deff effective diffusion coefficient [m 2 s21] and temperature may be expressed by sorption
isotherms or sorption isopsychrens (i.e., lines
Finally, to calculate the total flux j i of com-
of equal relative humidity). Apparently, the
ponent i, the diffusive flux may be superim-
most comprehensive set of such data available
posed on the convective flux:
in the literature for conditions including those
ji 5 jid 1 j ci i 5 v, a (3) above 1008C and vapor pressures above am-
WOOD AND FIBER SCIENCE
Monday Jun 09 2003 10:55 AM wood 35_403 Mp_459
Allen Press • DTPro System
File # 03em
bience are the sorption isopsychrens presented elastic material may be divided into elastic,
by Engelhardt (1979), who supplemented his delayed-elastic, and viscous components.
own measurements in the range between 110 However, irreversible changes of cell wall and
and 1708C with measurements reported by mat structure that happen instantaneously
Weichert (1963) for temperatures below upon loading are the dominant type of defor-
1008C. The curves presented by Engelhardt mation during densification of wood-furnish
were extrapolated for temperatures up to materials. The Burgers model does not repre-
2208C by fitting a second order polynomial to sent this type of deformation. To account for
the measured data, and were then represented both the visco-elastic material behavior and
in a look-up table with the relative humidity the instantaneous but irreversible deformation,
as dependent variable. Ren (1991) working with Humphrey supple-
Energy associated with the transition of wa- mented the Burgers model by adding a plastic
ter from the bound to vapor states, or vice ver- and micro-fracture (PMF) element in series
sa, is composed of the latent heat of sorption (Fig. 1a). A schematic of the stress-strain re-
(from the bound to the liquid state) and the lationship as it may be described by this rhe-
latent heat of evaporation (from the liquid to ological modeling approach is displayed in
the gaseous state). In the present model, a for- Fig. 1b. According to Ren’s (1991) definition,
mulation presented by Humphrey and Bolton strain caused by PMF effects occurs only if
(1989a) is used to relate the latent heat of the effective stresses exceed the yield strength
sorption and evaporation to moisture content of the microstructure; such deformation is not
(MC) and temperature. recoverable when the load is removed. The
Physical properties of the mat and the gases PMF element may be symbolized by a spring
within may vary with respect to both space that operates only in one direction.
Each of the five elements of the enhanced
and time. Most properties are highly depen-
Burgers model (hereafter referred to as the
dent on the local conditions of the mat. In ad-
Burgers-Humphrey model) presented in Fig.
dition, variations of the transport properties
1a can be described by a rheological coeffi-
(e.g., permeability, thermal conductivity, and
cient. These coefficients must be expressed as
obstruction factor) in the three principal direc- functions of temperature, MC, and density.
tions are accounted for in the model. A com- Given the rheological coefficients and the ef-
plete list of these dependencies is given in fective stress, the local deformation of the
Thoemen and Humphrey (2003a). wood furnish can be computed as the sum of
the four individual strain components.
Densification and the development of Two sets of rheological coefficients for the
internal stresses Burgers-Humphrey model are available so far,
either of which may be used as input data for
The wood furnish undergoes both instanta-
the simulation. The first was measured by Ren
neous and time-dependent deformation upon (1991) on fiber materials and fitted to equa-
application of pressing load. Such material be- tions by Haselein (1998). The second set has
havior may be illustrated and mathematically been derived from data presented by von Haas
described by employing rheological models in (1998) for fiber, particle, and strand materials;
which individual strain components are rep- those coefficients for fiber materials will be
resented by simple elements (e.g., a spring or used for the simulation run presented in this
a dashpot) or by a combination of such ele- paper.
ments.
The four-element Burgers model is fre- Numerical procedure and model
quently used in rheology to describe visco- implementation
elastic material behavior. The strain that de- The heat and mass transfer model described
velops when stress is applied to a pure visco- above consists of a set of constitutive flux
WOOD AND FIBER SCIENCE
Monday Jun 09 2003 10:55 AM wood 35_403 Mp_460
Allen Press • DTPro System
File # 03em
FIG. 1. (a) Five-element Burgers-Humphrey model, representing four strain components. E 5 modulus of elasticity,
V 5 viscosity, Epmf 5 coefficient for plastic and micro-fracture element. The non-linear behavior of each element is
not symbolized in this schematic. Refined after Ren (1991). (b) The four strain components described by the Burgers-
Humphrey model as a function of stress s and time t. e 5 strain.
equations (Eqs. (1), (2) and (4)) that are cou- MODELING THE CONTINUOUS PRESS
pled by local energy and mass balances. Fol-
Characteristics of the continuous
lowing this approach, the derivation of heat pressing process
and mass conservation equations and their
subsequent discretization was obviated. The earliest continuously working hot press
The wood-furnish mat to be modelled is for particleboard, the so-called Bartrev press,
subdivided into finite-differencing control re- was developed during the 1950s in England.
gions. Computations for each time step in- This first generation of continuous press was
clude three distinct consecutive procedures. not, however, very successful. New technolo-
First, heat and mass fluxes between the mid- gies developed during the 1970s and 1980s led
points of adjacent control regions are calcu- to today’s continuous presses that are larger
and more effective. Such presses for medium
lated under steady-state conditions to deter-
density fiberboard (MDF) typically have a
mine the diffusant quantities (e.g., energy)
production width of 2.10 to 3.15 m, press
flowing across the interfaces between neigh-
lengths up to 50 m, and production capacities
boring regions. Second, knowing gains and as great as 280,000 m3 per year (from data
losses of the diffusant within each control re- published by Sunds Defibrator 1998).
gion, new values for the state variables (e.g., The adhesive-treated wood-furnish mat is
temperature) are calculated. Finally, the state conveyed horizontally through the press be-
variables are corrected for compression or ex- tween two endless steel belts. For mat densi-
pansion of the control regions. To improve the fication and adhesive cure, heat and pressure
efficiency of computations, an adaptive time have to be transferred into the moving mat. A
step scheme in combination with a mechanis- simplified representation of the vertical cross-
tic implicit approach for the cross-sectional sectional structure of the press in the feed di-
convective flow calculations was adopted. The rection is given in Fig. 2. The rolling elements
in-house programming code for the simulation (usually steel rods or roller carpets), located
program is written in ANSI C. between the heating platens and the endless
WOOD AND FIBER SCIENCE
Monday Jun 09 2003 10:55 AM wood 35_403 Mp_461
Allen Press • DTPro System
File # 03em
System boundaries
Press designs vary among press manufac-
turers. In order to provide a model that may
be used to describe a wide range of different
press designs and methods of operation, a
rather adaptable formulation of the boundary
conditions with relatively straightforward ge-
ometries has been developed.
FIG. 3. Example of an in-feed section for a typical The scheme of the mat used in the present
continuous press. model is shown in Fig. 4. The horizontal sur-
WOOD AND FIBER SCIENCE
Monday Jun 09 2003 10:55 AM wood 35_403 Mp_462
Allen Press • DTPro System
File # 03em
the x-direction at these particular positions. On measurements can be found in Thoemen (2000).
the other hand, in order to avoid unacceptably
long execution times, the length of the grid
cells cannot be infinitely small. Clustering the
regions in the x-direction around the reference
regions, as shown in Fig. 6, minimizes the er-
ror caused by the discontinuity in the bound-
ary condition while keeping the number of
control regions reasonably small. This ap-
proach was justified by test runs with different
grid schemes. Other grid schemes may also be
specified for simulation runs.
FIG. 7. Specified pressing pressure and steel belt tem-
perature for the model demonstration simulation run. In
A SIMULATION EXAMPLE
the in-feed section, the pressing pressure cannot be de-
Initial and boundary conditions for the sim- fined directly, but is dictated by the radius and absolute
ulation run presented below were chosen to level of the upper steel belt.
WOOD AND FIBER SCIENCE
Monday Jun 09 2003 10:55 AM wood 35_403 Mp_464
Allen Press • DTPro System
File # 03em
FIG. 8. Predicted total gas pressure distribution within FIG. 9. Core gas pressures (l) and temperatures (.—
the central layer of the mat. Gas pressures are given as surface, m—core) measured in an MDF mat as it passes
absolute values. through the 28 m industrial press corresponding to the
presented model data (from Steffen et al. 1999).
(Steffen et al. 1999) represent the exact steel
belt temperature. In the absence of reliable in- trapped inside the mat escapes through the
dustrial data, the longitudinal temperature pro- edges; the gas pressure in the center slowly
file (Fig. 7) was estimated. This estimate is drops to an intermediate minimum of about 40
based on the steel belt temperature measured kPa above ambient. This relatively high level
right in front of the press entry (1328C) and of the first maximum and the subsequent in-
the temperature measured near the surface of termediate minimum corresponds well in trend
the mat. with the gas readings made on a corresponding
A 10-meter long mat segment was consid- industrial press (Fig. 9). Details of the mea-
ered for modeling purposes. The mat was as- surement methods are to be found in Steffen
sumed to be symmetrical along the width and et al. (1999).
the thickness, so that calculations were done Towards the end of the press, total gas pres-
for only one fourth of the mat. A modeling sure rises considerably. The generation of wa-
grid of 15 3 5 3 10 control regions (x, y, z) ter vapor, together with the compression of the
was used, and the control regions were clus- gas during the second densification step near
tered in the feed direction around the reference the end of the press, causes the final gas pres-
regions (Fig. 6). The size ratio of outermost to sure to reach a maximum of almost 150 kPa
innermost regions was seven. All results are above atmospheric pressure right before the
plotted for the reference regions, i.e., half way panel leaves the press. Throughout the entire
between the two boundaries perpendicular to pressing process, a near-parabolic gas pressure
the feed direction. The simulation run was ex- distribution is generated across the width of
ecuted with a specified minimum of 500 time the mat.
steps per second. Figure 10 shows the core layer distribution
of total gas pressure for the first 6 m of the
Model predictions press. The arrows represent the horizontal gas
Gas pressure predictions.—More than 95% velocity in the central layer, defined as the vol-
of the total gas pressure in the uncompressed umetric gas flow/sec/unit cross-sectional area. 2
mat before it enters the press can be attributed The gas pressure maximum is reached three
to air in the vapor-air mixture. The fast com- meters into the press, in the middle of the mat.
pression of the air leads to a predicted total Most of the vapor-air mixture escapes through
gas pressure maximum about 75 kPa above the edges in the in-feed section and through
atmospheric pressure early in the pressing cy-
2 The contact points of the mesh lines in Fig. 8 and the
cle (Fig. 8). During the subsequent period of
vector bases in Fig. 10 represent the centers of the grid
the pressing cycle, when the mat thickness regions. Hence, the first and last plotted values over the
stays almost constant, some of the gas en- width represent the values 5% from the actual mat edge.
WOOD AND FIBER SCIENCE
Monday Jun 09 2003 10:55 AM wood 35_403 Mp_465
Allen Press • DTPro System
File # 03em
acteristics, the mat expands again somewhat output and displayed for any location, and the
during the following stage of low pressure. model may be run with a diversity of raw ma-
If no further densification were to take terial and process configurations. Exploration
place, the density profile of the final panel of gas flow and heat transfer patterns, as well
would be relatively simple, with high densities as sensitivity studies, will improve our knowl-
near but not at the surfaces and a near-flat den- edge of the process. Such analysis will be pre-
sity distribution across the core. However, as sented in future publications.
specified for the simulation, the mat sustains
further densification in a second step near the ACKNOWLEDGMENTS
end of the press. At this stage, the cross-sec-
tional distributions of rheological properties The authors are grateful to G. Siempelkamp
differ greatly from those near the press entry. GmbH & Co. for facillitating our access to
Densification is greatest in the layers midway continuous presses for study. This work was
between surface and core, but almost no den- supported by USDA funds through the Center
sification occurs in the surface layers. A rather for Wood Utilization at Oregon State Univer-
complex density profile is thereby generated, sity and by the Federal Ministry of Economics
with intermediate density peaks in the former and Technology (Germany) through the Ger-
layers. Such intermediate density peaks have man Wood Research Foundation (DGfH). The
been measured frequently (e.g., Wang et al. authors also acknowledge the encouragement
2000). After leaving the press, the panel re- of Drs. A. Fruehwald and A. Steffen.
laxes, leading to a predicted spring-back of
about 4%. The core regions contribute most to REFERENCES
this spring-back, whereas the surface layers do
BOLTON, A. J., AND P. E. HUMPHREY. 1988. The hot press-
not expand significantly. The final panel thick- ing of dry-formed wood-based composites. Part I. A
ness is 20.4 mm. review of the literature, identifying the primary physical
processes and the nature of their interaction. Holz-
forschung 42:403–406.
CONCLUSIONS
DENISOV, O. B., P. P. ANISOV, AND P. E. ZUBAN. 1975.
The model presented in this paper is directly Untersuchung der Permeabilität von Spanvliesen. Holz-
technologie 16(1):10–14.
applicable to continuous presses. It accounts
ENGELHARDT, F. 1979. Untersuchungen über die Wasser-
for those features that are specific to this press dampfsorption durch Buchenholz im Temperaturbereich
type, including changing mat thickness and von 1108C bis 1708C. Holz Roh- Werkst. 37:99–112.
steel belt temperatures in the feed direction, GEFAHRT, J. 1977. Zur Spänevorwärmung mit Hochfre-
and the escape of vapor and air through the quenzenergie—Modell zur Berech-nung des Tempera-
turverlaufes in Vliesmitte bei der Heißpressung. Holz
horizontal surfaces immediately in front of
Roh-Werkst. 35:183–188.
and behind the press. Additional characteris- HASELEIN, C. R. 1998. Numerical simulation of pressing
tics of continuous presses, such as the possi- wood-fiber composites. Ph.D. thesis, Oregon State
bility to vary the mat thickness across the Univ., Corvallis, OR. 244 pp.
width of the press, are easy to incorporate into HUMPHREY, P. E. 1997. Thermoplastic characteristics of
the existing model. partially cured thermosetting-to-wood bonds. Pages
366–373 in Kajita and Tsunoda, eds. Proc. Third Pacific
The main purpose of this paper was to de- Rim Bio-Based Composites Symp., Kyoto, Japan.
scribe the strategy developed and implemented Wood Research Inst., Kyoto, Japan.
to simulate continuous presses, and to illus- . 1999. Bonding speed of adhesives: an automated
trate the potential of the model by displaying evaluation system. Pages 139–146 in M. Wolcott, ed.
and discussing results of just one simulation Proc. 33rd International Particleboard and Composite
Materials Symp., Washington State Univ., Pullman,
run. Clearly, the results presented cover only WA.
a small fraction of the range of possibilities. , AND A. J. BOLTON. 1989a. The hot pressing of
All variables included in the model may be dry-formed wood-based composites. Part II. A simula-
WOOD AND FIBER SCIENCE
Monday Jun 09 2003 10:55 AM wood 35_403 Mp_468
Allen Press • DTPro System
File # 03em
tion model for heat and moisture transfer. Holzforsch- MDF-mats during industrial continuous hot pressing.
ung 43(3):199–206. Holz Roh-Werkst. 57:154–155.
, AND A. J. BOLTON. 1989b. The hot pressing of THOEMEN, H. 2000. Modeling the physical processes in
dry-formed wood-based composites. Part VI. The im- natural fiber composites during batch and continuous
portance of stresses in the compressed mattress and pressing. Ph.D. thesis, Oregon State Univ., Corvallis,
their relevance to the minimization of pressing time, and OR. 187 pp.
the variability of board properties. Holzforschung 43(6): , AND P. E. HUMPHREY. 2003a. The physical pro-
406–410. cesses relevant during hot pressing of wood-based com-
, AND H. THOEMEN. 2000. The continuous pressing posites. Part I. Model description: Mass and heat trans-
of wood-based panels: a simulation model, input data fer. In preparation.
and typical results. Pages 303–311 in P. D. Evans, ed. , AND . 2003b. The physical processes rel-
evant during hot pressing of wood-based composites.
Proc. Pacific Rim BioBased Composites Symp. Can-
Part II. Model description: Rheology. In preparation.
berra, Australia. Australia National University.
VON HAAS, G. 1998. Untersuchungen zur Heißpressung
KAVVOURAS, P. K. 1977. Fundamental process variables in
von Holzwerkstoffmatten unter besonderer Berücksicht-
particleboard manufacture. Ph.D. thesis, Univ. of Wales, igung des Verdichtungsverhaltens, der Permeabilität,
UK. 156 pp. der Temperaturleitfähigkeit und der Sorptionsgeschwin-
KIM, J-W. 2003. Gas injection consolidation of miniature digkeit. Dissertation, Univ. of Hamburg, Germany. 264
natural fiber beams. Ph.D. thesis, Oregon State Univ., pp.
Corvallis, OR. 219 pp. WANG, S., P. M. WINISTORFER, W. W. MOSCHLER, AND C.
REN, S. 1991. Thermo-hygro rheological behavior of ma- HELTON. 2000. Hot-pressing of oriented strandboard by
terials used in the manufacture of wood-based compos- step-closure. Forest Prod. J. 50(3):28–34.
ites. Ph.D. thesis, Oregon State Univ., Corvallis, OR. WEICHERT, L. 1963. Untersuchungen Uber das Sorptions-
226 pp. und Quellungsverhalten von Fichte, Buche und Buchen-
STEFFEN, A., G. VON HAAS, A. RAPP, P. E. HUMPHREY, AND Preissvollholz bei Temperaturen zwischen 208C und
H. THÖMEN. 1999. Temperature and gas pressure in 1008C. Holz Roh- Werkst. 21:290–300.