PAPER PHYSICS

Flow conditions in the grooves of a Low-Consistency refiner

Lisa Prahl Wittberg, Magnus Björkman, Gohar Khokhar, Ulla-Britt Mohlin and Anders Dahlkild
KEYWORDS: Disc refiner, Chemical pulps, Fiber Kondora and Asendrych (2009) using CFD to
suspension, Flow, Softwood pulps, Hardwood pulps, numerically simulate the flow in a LC-refiner.
High-speed imaging, Computational Fluid Dynamics Fox et al. (1981) reported that the flow in a refiner is
directed inwards (towards the center) in the stator
SUMMARY: The flow pattern in the grooves plays a grooves and outwards in the rotor grooves. A feature also
major role for the homogeneity of refining as well as for observed by Herbert and Marsh (1968). Moreover, Fox et
the transfer and loading of fiber flocs in refining position al. (1982) found a circular/spiraling fluid motion within
on the bar edges. However, it is an area where very little the grooves due to the velocity difference between the
information is available. In the present study, flow stator and rotor. Similar results were presented by
conditions in the grooves in a Low-Consistency (LC) - Lumiainen (1994) for water flow in a conical refiner.
disc refiner were studied both experimentally and Another flow feature was observed by Lumiainen (1994)
numerically. The experimental study involved high-speed and Halme (1962) where a stagnation point was found to
imaging through a 3 cm peephole into a commercial occur in the stator groove in a conical refiner.
refiner. The Computational Fluid Dynamics (CFD) Regarding the rheology of a fiber suspension related to
simulation focused on the flow condition in a radial papermaking, review articles addressing this matter have
groove, considering both Newtonian and non-Newtonian been published during the past year (Derakhshandeh et al.
flows. Flow conditions for stator and rotor grooves were 2011; Lundell et al. 2011; Hämäläinen et al. 2011).
modeled along the groove at different angular speeds and Another often cited review article on fiber suspension is
pressure differences over the refiner. Both the that of Petrie (1999). Detailed studies of the fiber inter-
experimental and the modeling results show a dual flow action, with both neighboring fibers as well as with the
pattern in the grooves; a rotational/spiral movement at the flow is another vivid research area (Krochak et al. 2009;
top of the groove and a flow in the direction of the groove Tornberg Shelly 2004; Carlsson et al. 2011). In terms of
at the bottom, which to the authors knowledge has not increasing the understanding of the underlying physical
been reported in literature. The strong vortical motion at properties governing these flows, this is an important
the top of the grooves observed both for the rotor and the field.
stator are believed to be important for placing the fibers Fiber suspensions are commonly classified as a non-
onto the bar edges and to induce shear forces in such a Newtonian fluid, i.e. the shear stress and shear rate
way that the fibers get treated. Moreover, a large exhibit a non-linear behavior leading to that the viscosity
sensitivity to suspension properties in terms of the of the suspension cannot be characterized by a constant
development of flow pattern was detected. viscosity (Hemström et al. 1976; Lee Duffy 1976).
ADDRESSES OF THE AUTHORS: Lisa Prahl However, in a recent study by Chaussy et al. (2011),
Wittberg ([email protected]), Linné FLOW center, investigating the rheological properties of a 4% bleach
KTH Mechanics, Royal Institute of Technology, SE-100 softwood fiber suspension during refining conditions, it
44 Stockholm, Sweden. Magnus Björkman was found that for values below a critical shear rate, the
([email protected]), Innventia AB, Box pulp displayed Newtonian behavior.
5604, SE-114 86 Stockholm, Sweden. Gohar Khokhar In literature, great focus has been directed towards
([email protected]), KTH Mechanics, Royal Institute of Thermomechanical pulping (TMP), an area to which the
Technology, SE-100 44 Stockholm, Sweden. Ulla-Britt studies by Huhtanen (2004) and Lindstedt et al. (2009)
Mohlin ([email protected]), Innventia AB, are connected to. Although literature displays a lot of
Box 5604, SE-114 86 Stockholm, Sweden. Anders research concerning the refining process (Hietanen,
Dahlkild ([email protected]), Linné FLOW center, KTH Ebeling 1990; Roux 2001; Sjöström 1993, Heymer et al.
Mechanics, Royal Institute of Technology, SE-100 44 2011, Muhic et al. 2011), many questions still remain to
Stockholm, Sweden. be answered concerning LC-refining. There are issues
Corresponding author: Lisa Prahl Wittberg concerning the effect the retention time in the refiner has
In papermaking, refining is used to control the paper on the refining results that need to be resolved in order to
quality. Refining is carried out in disc or conical refiners understand LC-refining. Ryti and Arjas (1968) carried out
at 3-5% consistency. The flow conditions in the refiner experiments regarding retention time and its effect on
are important for the refining action. pulp quality. Moreover, the flow pattern in the groove
One of the most cited work concerning the flow in a can be expected to play a major role for the homogeneity
refiner was carried out in the beginning of the 1980's by of refining and for the transfer of fiber flocs into refining
Fox et al. (1981, 1982), using high speed imaging for positions on the bar edges. Thus, motivating the present
flow measurements in a 12" (305 mm) Plexiglas disc study in which both visual observations through high-
refiner. However, the consistency of the pulp was fairly speed camera recordings and numerical simulations using
low, 0.1% and the refiner was run with a wide gap CFD have been applied in order to further increase the
clearance. Lumiainen (1994) used Laser Doppler knowledge of the flow features occurring in the stator and
Anemometry (LDA) to study the flow in the stator rotor grooves of a LC-refiner. The purpose of the
grooves using water. A recent study was performed by numerical simulations was to aid the interpretation of the

Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 173

PAPER PHYSICS

Fig 1. The experimental set-up including the gap sensor hole

used to capturing the flow.
experimental results as well as extend the description of
the flow beyond what could be observed.

Materials and Methods

In this study, both experiments and numerical simulations Fig 2. The Plexiglas dummy used in the experiments, with a
have been carried out. Details regarding the set-ups are diameter of 33 mm. The width of the grooves and bars visible in
given in the following section. picture is 4.7 mm.

Experimental set-up order to study the flow conditions in the bottom of the
stator grooves, Particle Image Velocimetry (PIV) was
In the experiments, a 24'' (600 mm) Double Disc (DD)-
applied, using a Phantom V5.1 high-speed camera.
refiner (Beloit Jones) was used. Two fillings were
Optical access was obtained through a Plexiglas dummy
applied, a coarser softwood filling of type
mounted in a 33 mm wide hole, originally made for a gap
24901/24902BJ and a finer softwood/mixed furnish clearance sensor. The Plexiglas dummy was tailored to fit
filling, type 24101/24102BJ (National Refiner Plate). Bar
the pattern of the bars and grooves in the filling. Figs 2 –
geometries for the coarser filling were; bar width 4.0 mm,
3 display the Plexiglas dummy used in Trial 1 as well as
groove width 4.7 mm and groove depth 7.1 mm and, for
the position of the Plexiglas dummy in the stator disc
the finer filling the dimensions were 3.2 mm, 3.2 mm and
(indicated by the arrow). As shown in Fig 1, illumination
7.1 mm respectively. The trials were performed at three
was obtained from a ring of 9 halogen lamps positioned
occasions. In Trial 1 and 2, the coarser filling was used
by the Plexiglas dummy. The illumination was in the
whereas in Trial 3 the finer filling was considered. An
order of 2000 lumen. In Trial 1, a mirror was inserted in
overview of the refining conditions is given in Table 1.
one of the bars in the Plexiglas dummy to allow studying
The furnishes used in Trial 3 were bleached softwood
the flow conditions around the bar edge and at about 2.5
kraft pulp, evaluating three different consistencies (2, 3,
mm down into the groove. Fine grinded coffee particles
4%) and mixtures (100/0, 50/50, 75/25, 90/10 and 0/100)
were used as tracer particles. By recording the movement
of unbleached softwood and bleached eucalyptus pulp at
of the tracer particles, the velocities were computed and
4% consistency.
related to the fiber flow.
High-speed imaging
The high-speed imaging set-up is displayed in Fig 1. In

Table 1. Overview of the refining conditions.

Trial 1 Trial 2 Trail 3
Filling 901/902 901/902 101/102
Pulp BSK BSK, BHK BSK, BHK, mix
Pulp consistency % 3.5 4 2, 3 and 4
Refiner speed rpm 600, 750 600, 750, 900 750
Net flow rates lit/min 600 500, 800, 1200 600
Power applied kW 90, 140 70 - 220 130
Exposure time s 20, 70 13 30
Frame rate fps 4300 8300 5706
No of experiments 3 12 9
174 Nordic Pulp and Paper Research Journal Vol 27 no.2/2012
 PAPER PHYSICS

Fig 4. The two geometrical set-ups used in the numerical

simulations; referred to as Case 1 and Case 2. A back pressure
Fig 3. The stator disc in a 24'' (610 mm) DD-refiner (Beloit is applied over the groove, i.e. low pressure at the centre of the
Jones) with mounted fillings. The hole in which the Plexiglas refiner as compared to the periphery.
dummy is mounted is marked by the arrow.
Table 2. The parameters included in the Herschel-Bulkley
model that were tested.
Consistency factor K [Pasn] 0.01, 0.1, 1 and 10
Yield stress y [Pa] 50, 100, 200 and 300
Shear index n[-] 0.1, 0.5 and 0.9
Numerical modeling and Geometry
For the CFD simulations, Ansys Academic Research, Fig 5. A cross-section of the geometry, showing the boundary
release 12.1 FLUENT solver was used. The RNG k -  conditions
turbulence model was applied (Yakhot et al. 1992), using of the groove was 125 mm. The origin was located 175
the SIMPLE scheme for the pressure-velocity coupling mm from the center of rotation. The gap clearance
with second and first order discretization in space and between the groove and the flat plate was 0.2 mm.
time, respectively. The computational mesh was created in Ansys Academic
The fiber suspension was modeled as a single-phase Research, release 12.1 ICEM. Three different grid
(individual fibers were not modeled) and incompressible resolutions were tested; consisting of 284 031, 562 518
fluid, considering both Newtonian and non-Newtonian and 806 157 nodes. The grid sensitivity showed that a
formulations. grid of 284 031 was sufficient in order to properly resolve
In the Newtonian simulations, the fluid density was set the flow field.
to 998 kg/m3 whereas four different fluid viscosities (µ)
were investigated, 0.001, 0.01, 0.1 and 1 Pas (µ water = Boundary Conditions and Case set-up
0.001 Pas), in order to simulate more viscous fluids. Depending on the boundary conditions applied, the flow
In the non-Newtonian simulations, a Herschel-Bulkley in the groove model represents either the flow in a stator
model was applied, accounting for both the yield stress or in a rotor, with the assumption that the top wall is a flat
and shear-thinning properties of a suspension to compute plate with a different rotational speed instead of another
the apparent viscosity (ap) of the suspension. The model groove model. Moreover, the groove was aligned in the
parameters that needs to be assigned are the consistency radial direction, i.e. without an angle as in the actual
factor (K), the yield stress (y) and the shear index (n). refiner. No-slip conditions were applied at the walls and
However, the numerical values that are to be appointed to periodic boundary condition was applied as showed in
these parameters are not fully resolved (Hemström et al. Fig 5.
1976; Wikström 2002; Hammarström 2004). Therefore, a In order to simulate the flow in the stator, back pressure
sensitivity study was carried out. The values tested are differences (pback) of similar magnitudes as found in the
displayed in Table 2. experiments were applied over the inlet and outlet
Instead of considering the flow in the whole refiner, the boundaries to impose an inward flow. The flat plate in
geometry studied in the numerical simulations was Case 1, see Fig 4, was given a constant velocity of 18.6
represented by a single groove. In a disc refiner the stator m/s, corresponding to 750 rpm at the center of the
and rotor grooves are facing one another. However, in the groove, over the whole length of the groove. For Case 1,
numerical simulations, the geometry was further the effect of different dynamic viscosities was studied,
simplified by modeling a groove and a flat plate at considering a Newtonian fluid behavior. Also, a non-
different rotational speeds, Fig 4. The dimensions were Newtonian fluid flow case was investigated. The different
the same as in the experimental set-up; the groove width viscosities investigated in the Newtonian simulations
and depth were 4.7 and 7.1 mm, respectively. The length were 0.001, 0.01, 0.1 and 1 Pas. Moreover, for these
Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 175
PAPER PHYSICS

simulations, the pressure differences studied were 100, fibers as observed through the mirror placed in the
150 and 200 kPa. In the non-Newtonian flow case, the bottom of the groove. The shaded area visible in the
pressure difference was set to 150 kPa, defining the background is passing rotor bars and grooves. The flow
consistency factor, the yield stress and the shear index as speed changed with the refiner speed and throughput. The
shown in Table 2. rotation pattern found in Trial 1 was not possible to
In Case 2, the flow pattern along the groove in both observe although it has to appear at the top of the groove.
stator and rotor was investigated by applying a rotational The difference in flow pattern between the Trial 1 and 2
velocity to the flat plate. Here, solely a Newtonian fluid for the softwood pulps was unexpected, as a non-
was considered. The viscosity was set to 0.1 Pas with the spiraling flow in the direction of the groove has not been
groove positioned 175 mm away from the center of previously reported in the literature. The only difference
rotation in order to consider the offset in radial position between the trials was the slightly lower consistency used
of the grooves occurring in the refiner used in the in Trial 1.
experiments. The rotational velocities () studied were Moreover, Trial 2 also included two experiments using
600, 750 and 900 rpm. The pressure differences applied a hardwood pulp of 4% consistency. The fillings were
over the groove were 50, 75, 100 and 125 kPa. The much too coarse to be used together with hardwood
reason for lower pressure differences as compared to pulps. However, the results are still of interest as the flow
Case 1 is due to matching the flow in the rotor and the pattern in the grooves was completely different from
stator (the flow magnitude in the rotor is smaller than in what was observed for the softwood pulp. In this case, the
the stator). rotational flow extended all the way down to the bottom
To simulate the flow in the rotor, a rotational framework of the groove; see Fig 6(bottom). The reason for this
was used to model the rotation of the groove, i.e. the flat different flow pattern for the hardwood pulp was
plate was given a relative angular velocity. Therefore, in hypothesized to be an effect of the difference in floc size
the figures showing the simulation results of the rotor, the in relation to the dimensions of the grooves.
direction of the rotational motion of the flat plate will be Trial 3 was run with finer fillings, 3.2 mm wide grooves
indicated. instead of 4.7 mm as in Trial 1 and 2, and included
experiments with a bleached softwood pulp at three
Results consistencies, 2, 3 and 4%. The 4% consistency showed
In this section, the experimental results will first be an unexpected low speed that was believed to be due to
discussed, followed by the results from the numerical clogging of the groove. For 2% and 3% consistency, the
simulations. The main purpose with the numerical flow in the direction of the groove was dominating.
simulations was to aid the interpretation of the However, at a consistency of 2%, a greater element of
experimental results and to obtain information regarding rotation was observed. Thus, this trial indicates that the
the flow features appearing in regions in which results consistency may have an effect on the flow pattern in the
could not be obtained by the experimental studies, i.e. grooves favoring rotation extending all the way to the
along the groove and in the rotor grooves. bottom of the groove at lower consistency.
From the results obtained in Trial 2, it was also
Flow behaviour as revealed by the experiments hypothesized that floc size in relation to groove width has
The experimental results revealed a complex flow in the an impact on the level of rotational movement that can be
stator grooves. The difference in speed between the rotor observed at the bottom of the groove. With the finer
and the stator induce a rotational motion of the fiber
suspension. This rotation was in turn mixed with a
transport of the suspension moving towards the center of
the refiner. Depending on the suspension properties and
the running conditions of the refiner (speed and
throughput), several different flow situations occurred.
In Trial 1, using a bleached softwood kraft pulp, BSK, a
rotational movement, as reported by Fox et al. (1982),
was dominant for a refiner speed of 600 rpm. The
rotation was observed both at the bottom of the groove
and at the topside of the groove, with the speeds being
somewhat higher at the top of the groove as compared to
the bottom. The rotation was at an angle to the direction
of the groove. For 750 rpm the flow pattern at the bottom
of the groove was a mix between rotation and a flow in
the direction of the groove towards the center of the
refiner. Fig 6. Trajectories of the particles showing the direction of flow
Trial 2 was carried out at a slightly higher consistency at the bottom of a groove as observed in the mirror for refining
and included more combinations of refiner speed and at 750 rpm and 800 lit/min (top) with softwood pulp at 4%
throughput. In this case, the flow in the bottom of the consistency and for hardwood pulp at 500 lit/min and 750 rpm
groove was mainly in the direction of the groove for all (bottom). The different colours represent different particles. The
trials using softwood pulp, Fig 6(top). Fig 6 shows the main motion is from left to right i.e. towards the centre of the
refiner.
176 Nordic Pulp and Paper Research Journal Vol 27 no.2/2012
 PAPER PHYSICS

filling, a series of experiments was performed with

mixtures of unbleached softwood pulp and bleached
eucalyptus of 4% consistency. Also for these mixtures,
the flow in the direction of the groove was the dominant
flow. However, the element of rotational movement that
could be observed at the bottom of the groove increased
with increasing eucalypt content in the mix. This is in
line with the hypothesis that the floc size in relation to the
groove size has an impact on how far down into the
groove the rotational movement will extend.
The different flow patterns observed for the softwood
and the hardwood pulp with the coarse filling are
believed to be due to difference in fiber length/floc size.
The flow pattern observed for hardwood pulp is similar to
that earlier reported by Fox et al. (1981) and Lumiainen
(1994). In these studies, 0.1% fiber suspension and water
was used for the flow measurements, respectively. The
low viscosity seems to favor the flow pattern where the
rotation extends the whole depth of the groove.
The flow in the direction along the groove, as found in
Fig 7. The velocity versus pressure difference in the stator for
this study for the softwood pulp and also for all the
different net flow (lit/min) and refiner speeds (rpm) for all
furnishes investigated using the finer filling, has not been
experimental sessions investigated. The results indicate a linear
reported in the literature. In this case, the fibers are
trend between the speeds in the direction of the groove versus
assumed to form fiber flocs of similar size as the groove
the pressure increase, included in the figure as a black line.
width. Therefore, the suspension will move as a single
body filling the entire width of the groove, but not the
entire groove depth.
The Newtonian flow case
The average transport flow speed was found to be
within the range of 2 – 8 m/s, although locally as high as Similar to what was observed in the experiments, the
15 m/s. A strong positive correlation was detected numerical simulations of the Newtonian fluid flow in the
between the pressure increase over the refiner and the stator groove indicated that the fluid properties have a
average return flow speed in the bottom of the stator clear effect on the flow in the bottom of the stator groove.
grooves, Fig 7. Similar speeds were observed for the two Figs 8 – 9 display the motion in the cross-section located
fillings at a given pressure increase of the refiner. The in the middle of the stator groove for different viscosities
transport flow speed increased with increased refiner for p = 150 kPa and  = 750 rpm, represented by
speed and was reduced with increasing net flow. streamlines and the velocity magnitude, respectively. The
As a result of these experiments, the flow pattern in the simulations showed that with a viscosity of 0.001 – 0.01
grooves gave rise new questions as well as increased Pas, a rotational motion was found in the entire groove,
interest, since it is easy to picture several impacts on the corresponding to what was observed by Fox et al. (1982)
operation of the refiner connected to changes in flow and Lumiainen (1994). For a viscosity of 0.1 Pas, two
conditions in the grooves. The main factors are the different zones appeared in the groove, a strong rotational
retention time for fibers in the refiner affecting the motion in the top of the groove and a weak rotation in the
homogeneity of fiber treatment, and the trapping of fibers bottom of the groove, Fig 8. The flow in the bottom zone
to the bar edges. The results from the experimental work was mainly in the direction of the groove. This
contained some unexpected flow patterns. Modeling was corresponds well with the experimental results for the
applied to assist in increasing the understanding how softwood.
these complex dual flows were obtained. The velocities found in the simulation using a viscosity
of 0.1 Pas corresponded well with the velocities of
Numerical modelling; Newtonian vs. non-Newtonian approximately 6 m/s detected in the bottom of the groove
In these simulations, the flow in a stator groove was in the experiments for 150 kPa, a net flow rate of 500
modeled (Case 1). The rotational motion of the flat plate lit/min and a rotor speed of 750 rpm, see Fig 7.
was neglected; instead the plate was given a constant Moreover, the viscosity of 0.1 Pas is of the same order of
velocity over the whole length of the plate corresponding magnitude as 0.37 Pas, reported by Radoslavova et al.
to 750 rpm at a radius of 237.5 mm. Both Newtonian and (1996) to correspond to a 4% fiber consistency.
non-Newtonian fluids were considered. Furthermore, Chaussy et al. (2011) pointed out that the
The pressure difference over the groove was found to apparent viscosity of the fiber suspension in the gap
only influence the velocity of the fluid and not the flow clearance in a refiner is approximately 100 times than
pattern (in the Newtonian flow case). Thus, only pback = that of suspensions of individualized fibers.
150 kPa is presented in this paragraph.

Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 177

PAPER PHYSICS

Fig 8. The surface streamlines in the cross-section located in the middle of the stator groove for different viscosities at pback = 150
kPa and  = 750 rpm. Starting from the left, the viscosities are as follows; 0.001, 0.01, 0.1 and 1 Pas.

Fig 9. The velocity magnitude in the cross-section located in the middle of the stator groove for different viscosities at pback = 150
kPa and  = 750 rpm. Starting from the left, the viscosities are as follows; 0.001, 0.01, 0.1 and 1 Pas.

Fig 10. The surface streamlines in the cross-section located in the middle of the groove applying a non-Newtonian model (the
Herschel-Bulkley model) for different consistency factors at pback = 150 kPa and  = 750 rpm. Starting from the left, the
consistency factors are as follows; 0.01, 0.1, 1 and 10 Pasn, n = 0.5 and y = 100 Pa.
The Non-Newtonian flow case for detailed simulations of these flows, such as the Lattice
As described in the Methods section, a Herschel-Bulkley Boltzmann method combined with an external forcing or
model was applied to include the non-Newtonian the Immersed Boundary method combined with a flow
behavior of a fiber suspension into the numerical solver (Wu and Aidun 2010; Peskin 2002). However,
simulations. However, it was ambiguous what numerical simulating fiber flows are far from trivial, and problems
values to use for the shear index, consistency factor and are encountered when increasing flow rates, fiber volume
yield stress in order to best model the flow. Therefore, fractions and modeling flexible fibers. Commonly,
simulations were carried out to evaluate these parameters, simulations are limited to simple geometries. The study
Table 2. by Wu and Aidun (2010) is one of few studies available
The simulations showed that the shear index and the focusing on the flow of flexible fibers of volume
consistency factor influenced the flow the most. fractions of 1.7 – 12.4%, considering pipe flow. In the
Considering the overall flow velocities, a shear index of present study, a more complex set-up is considered
0.9 and a consistency factor of 0.1 Pasn resulted in a flow including moving boundaries.
corresponding reasonably well with that measured in the According to the above sections, the results obtained by
experiments. the Newtonian model provided good agreement with the
In general, the non-Newtonian simulations experimental results. Furthermore, insight into the
overestimated the flow velocity as compared to the behavior of the suspension can also be obtained from the
experiments. Moreover, as displayed in Fig 10, the non- power and gap clearance relationships. Chaussy et al
Newtonian flow cases investigated failed to capture the (2011) suggested that the refiner has many similarities
feature of the motion along the bottom of the groove with a parallel plate rheometer. In the study by Chaussy
observed for softwood pulp, obtained in the simulations et al. (2011), a linear relationship between power and the
using a constant viscosity of 0.1 Pas. inverse of gap clearance was interpreted as an indicator of
In the refiner there exists a complex interaction between a Newtonian behavior of the fiber suspension and a
the fibers and the fluid. If performing a detailed nonlinear declining power versus 1/gap relationship was
numerical simulation of a fiber flow properly, the interpreted as a shear thinning behavior. Thus, for low
individual fibers need to be correctly represented along shear rates, as considered in the present study, the fiber
with the consideration of fiber-fiber and fiber-flow suspension was suggested to display a Newtonian
interactions. There are direct numerical methods available behavior. In the experiments performed in the present

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

 PAPER PHYSICS

study it was not possible to measure gap clearance as the

sensor position was used for filming. However, in other
similar experiments using the coarse fillings and bleached
softwood kraft pulp, a linear relationship was observed
for 3,6% and 800 lit/min in the power range 70 – 240
kW. For 400 lit/min, a nonlinear behavior was observed
above 160 kW, Fig 11. For the finer fillings linear
relationships were observed within the whole power
range 70 – 240 kW (4%; 500 and 1000 lit/min).
All of the above, combined with the uncertainty
followed by the use of non-Newtonian models in terms of
how to best tune the model, the simulations for Case 2
considering the flow in both the rotor and the stator was
carried out for a Newtonian fluid. In the following
sections, solely µ = 0.1 Pas is considered, a choice Fig 11. The effect versus 1/gap clearance using coarse fillings
supported by the study by Radoslavova et al. (1996). (as in Trial 3) and 3.6% bleach softwood kraft pulp for net flow
rates corresponding to 400 and 800 lit/min.
Case 2: The flow in the stator and the rotor applying a
viscosity of 0.1 Pas
In the above section we found good agreement between
experiments and numerical modeling for the stator. How-
ever, in the experiments it was not possible to study the
flow inside the rotor. Hence, numerical simulations were
used to evaluate and compare the flow in the stator and
the rotor, including the effect of varying radial speeds.
Fig 12 displays the mass flow rate through both stator
and rotor. It is clear that the flow in the stator is
independent of the angular speed, but increases with
pressure difference. However, the flow in the rotor
decreases with pressure difference and increases with
angular speed. In reality, a valve on the outlet from the
refiner is used to set a target net flow rate and this
controls the Δpback. The net flow rate is the difference Fig 12. The mass flow rate through the rotor (blue) and the
between the flow in the rotor and the stator, implying that stator (red) for different backpressure differences and angular
only conditions above the red line in Fig 12 are relevant speeds.
for the operation of the refiner. compared to further down in the groove. Moving further
The simulations showed that the flow pattern (velocity downstream in the groove, the vortical motion is
distribution) inside the groove was fairly independent of strengthened and at 90 mm, a second recirculation zone
flow speed. appears in the bottom of the groove. This secondary
In the following sections, the simulations will show that recirculation zone develops and becomes larger with
the flow pattern inside the stator and rotor grooves dis- decreasing distance to the center of the refiner.
play similar features. The applied backpressure difference Considering the top vortical motion, the rotational flow in
over the groove and the refiner speed were set to 50 kPa this area is not changing as much as the lower
and 750 rpm, respectively. recirculation zone. The center of the top recirculation
The Stator zone is shifted towards the left, corresponding to the
Fig 13 shows the flow in the stator following a tracer direction of the motion of the flat plate. Fig 14 indicates
particle. Driven by the pressure difference, the flow in the that the rotational motion in the top of the groove acts in
stator grooves are directed towards the center of the order to transport the fibers towards the bar edge.
refiner (negative y-direction), as found in the experi- The Rotor
ments. Moreover, in the top of the groove a vortical The flow in the rotor is affected by two forces acting in
motion is detected. A section of rotational motion is also opposite directions; the pressure difference over the
found in the bottom of the groove, although exhibiting groove acting in order to move the flow towards the
much less rotation as compared to the top of the groove. center of the refiner and the centrifugal forces acting in
The flow development within the stator groove at order to move the flow outwards. The resulting flow
different cross-sections along the length of the groove is direction is towards the periphery (positive y-direction)
depicted in Fig 14, showing the y-velocity component of as showed in Fig 15. Compared to Fig 13, the flow in the
the velocity vector in color with the arrows representing rotor appears to experience a lower spiraling frequency.
the xz-components. As the flow enters the stator, the Fig 16 shows the variation in the y-component of the
vortical motion has not had time to develop and is there- velocity vector and the rotation in terms of the xz-
fore less significant. However, due to the rotation of the components by color and arrows, respectively. The top
flat plate, the velocities near the plate are greater as
Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 179
PAPER PHYSICS

with increasing pressure difference in the rotor, and

increases with increasing pressure difference in the stator.
Furthermore, studying the streamlines for the rotor and
the stator groove a difference is observed. In the rotor,
three different recirculation zones are detected for 0.1
Pas,750 rpm and 50 kPa as compared to the two zones
found for the stator, Fig 18. One explanation may be due
Fig 13. Trace of a fluid particle motion within the stator groove to the presence of the Coriolis force in the rotor, leading
for = 0.1 Pas, pback = 50 kPa and  = 750 rpm. to the appearance of a counter-rotating vortex. Fig 18
also shows that the rotational component in the bottom of
the groove is less significant compared to the rotational
motion at the top.

Concluding Discussion
The flow in a LC-disc refiner has been investigated
through experiments and numerical simulations. In the
numerical simulations, the flow pattern at a viscosity of
0.1 Pas was found to correspond well with the
experiments. The viscosity of 0,1 Pas is also of the same
magnitude as what is reported in the literature for a pulp
Fig 14. Variation of the y - component of the velocity at different fiber suspension at 4% consistency (Radoslavova et al.
locations within the stator groove. The xz - velocity vectors are 1996). The modeling replicated the dual flow observed
represented by arrows. pback = 50 kPa,  = 0.1 Pas and  = experimentally; a strong rotation/spiraling movement at
750 rpm. The motion of the rotation of the flat plate is directed the top of the groove and a flow in the direction of the
to the left. The negative sign of the velocity indicates the groove with very little rotation at the bottom. The overall
direction that is towards the centre of the groove. movement between the spiraling and the radial movement
was in the same direction; inwards in the stator grooves
and outwards in the rotor grooves. In previous work by
Fox et al. (1981) and Lumiainen (1994) only the
rotational/spiraling movement was observed. Their
results better reflect what was modeled at lower
viscosities of 0.001 Pas (water (Lumiainen 1994)) and
0.01 Pas ( 0.1% consistency (Fox et al. 1981)). The only
other study reporting the dual flow pattern in the grooves
is the numerical study by Kondora and Asendrych
Fig 15. Trace of a fluid particle motion within the rotor groove (2009). However, they reported that the flow in the
for  = 0.1 Pas, pback = 50 kPa and  = 750 rpm. bottom (outwards) and the top (inwards) of the stator
recirculation zone increases as the distance to the center grooves had different directions, which in turn is
is increasing. This is expected as the rotational velocity dissimilar to our experimental and numerical results. This
(r) of the flat plate increases with distance to the center. difference in results is probably connected to the choice
However, at approximately 45 mm and onwards, the of boundary conditions. In the present study, the flow
recirculation zones remain close to constant. At this was only controlled by the pressure difference between
point, the secondary recirculation zone appearing in the the inlet and the outlet, allowing back flow to occur in
bottom of the groove is also observed. both ends of the stator grooves. Kondora and Asendrych
(2009) used a constant inlet flow rate boundary condition
Comparison of the Rotor and Stator that may control the flow direction.
Comparing the flow pattern in the stator and rotor The balance between the flow speed in the stator and the
grooves, similar trends were observed, displayed in Fig rotor grooves determines the retention time for the fiber
17, where the variation in the y-velocity component for suspension in the refiner and thus the probability of the
back pressure ratios of 50, 75 and 100 kPa for both the fibers getting mechanically treated. The rotor flow speed
stator and rotor for  = 900 rpm are shown. With was found to be reduced by an increase in pressure
increasing pressure difference over the groove, the difference and a reduction in angular velocity. The flow
velocity in the bottom of the groove decreases whereas at rate in the stator grooves was only affected by the
the top of the groove, the velocity increases. Also, the pressure difference, not by the angular speed. As changes
rotational motion is strengthened with an increase in in the net flow rate are usually activated by changing the
pressure difference. Moreover, the velocity magnitudes pressure difference, a change in the flow rate implies a
are greater for the rotor as compared to the stator. change in the level of recirculation inside the refiner and
However, as expected, regarding the average velocity the probability for fiber treatment.
magnitudes found in the grooves, the velocities decreases

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

 PAPER PHYSICS

Fig 16. Variation of the y - component of the velocity at different locations within the rotor groove. The xz - velocity vectors are
represented by arrows. pback = 50 kPa,  = 0.1 Pas and  = 750 rpm. The motion of the rotation of the flat plate is directed to the
left. The negative sign of the velocity indicates the direction that is towards the centre of the groove.

Fig 17. Variation of the y - component of the velocity for different pressure ratios; pback = 50, 75 and 100 kPa for the rotor (top) and
the stator (bottom). The xz - velocity vectors are represented by arrows. = 0.1 Pas and  = 900 rpm. The cross-sections are
located the middle of the groove (62.5 mm). The motion of the rotation of the flat plate is directed to the left.

Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 181

PAPER PHYSICS

This study, combining experimental studies with

numerical simulations has put a focus on the flow
conditions inside the refiner. It has shown the large
sensitivity to suspension properties in terms of the
development of the flow pattern. The pressure difference
over the refiner is often used as a criterion of how well
the refiner is operating. This study provides insight into
interpreting the pressure difference and the differences
developed when changing refining parameters.

Conclusions
The results obtained in this study through experimental
and numerical studies can be summarized as follows;
- An unexpected flow pattern was observed for softwood
pulp of 4% consistency. The experiments revealed a
dual flow pattern consisting of a strong
rotating/spiraling motion at the top of the groove and a
Fig 18. Vector plot (left) and streamlines (right) of the flow in a
flow with little rotation in the bottom of the groove.
rotor (top) and stator (bottom) groove taken in the midplane of
- The dual flow pattern was found to move inwards,
the groove (62.5 mm).  = 0.1 Pas, pback = 50 kPa and  = towards the center of the refiner, in the stator grooves.
750 rpm. The motion of the rotation of the flat plate is directed - The numerical simulations using a constant fluid
to the left. viscosity of 0.1 Pas reproduced the experimental
findings.
Moreover, the flow conditions at the top of the grooves - In the simulations, the dual flow pattern was found to
is believed to also play a major role in bringing the fibers move outwards in the rotor grooves.
into position to get treated and for the mechanical loading - The experiments and the numerical simulations
of the fibers (Mohlin 2010). The high speed films indicated that the consistency may affect the flow
indicated that the fibers are brought to the bar edge by the pattern observed in the grooves, favoring a rotational
rotational movement and when collected on the bar edge, motion that is extended to the bottom of the groove
the fibers get strained by two shear fields; one along the with decreasing consistency.
bar surface originating from the speed difference between - Decreasing the floc size in relation to groove width had
the rotor and stator, and another directed downwards into a similar effect as reducing the viscosity.
the groove originating from the rotational movement of - The pressure increase and the average return speed in
the fluid. The shear field along the bar surface is the bottom of the stator groove were found to strongly
controlled by the gap clearance and the shear field along correlate.
the groove side is defined by the development of the - In the simulations, the flow in the stator was found to
rotating/spiraling movement. only depend on the backpressure applied over the
This numerical study showed that the flow pattern in the groove, whereas the flow in the rotor groove showed
groove varies along the length of the groove. As shown in dependency on both the backpressure and the angular
Figs 14 and 16, the spiraling movement is not fully velocity assigned to the flat plate.
developed until about 30 mm into the groove. The - Comparing the flow pattern in the rotor and stator
positioning of fibers on the bar edge as well as the fiber grooves, the numerical simulations displayed the
loading is probably not achieved until some distance into occurrence of three different recirculation zones
the groove. Also, the magnitude of the fiber loading will appearing in the rotor grooves as compared to the two
increase along the radius of the refiner due to the increase zones found in the grooves of the stator.
in actual speed of the rotor. This is in turn is transferred
to an increased speed in the rotation/spiraling movement
at the top of the refiner.
The experimental study revealed very different results Acknowledgements
for the softwood and the hardwood pulps in coarse The crew at the Innventia pilot refining plant are thanked for the
fillings although using the same consistency. The realization of the trials, Leif Falk for making the Plexiglas
hardwood pulp behaved like a low viscosity pulp dummies, Magnus Hillergren for carrying out the high-speed
suspension; the rotation was clearly observed also in the imaging and Lars Thomson for analyzing the films. The mirror in
bottom of the groove. The difference between the two the Plexiglas dummy used in trial 1 was made possible by the
pulps is the floc size. This can be interpreted as that in assistance of Professor Alexander Grishin, KTH
order to develop the dual flow pattern it is important that The financial support from the Swedish Research Council
the size of the fiber flocs fit the groove and that there is (Vetenskapsrådet) and the Swedish Energy Agency
not room for more than one floc. Hardwood pulps are (Energimyndigheten) is greatly acknowledged as well as the
usually refined in finer fillings with a typical groove funding obtained by an industrial consortium: Billerud, Holmen,
width of 2 – 3 mm. Korsnäs, Mondi, Södra, UPM and Voith.

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

 PAPER PHYSICS

Literature Lumiainen, J. (1994): Plate pattern and fibre dimensions have

Carlsson, A., Håkansson, K.H., Kvick, M., Lundell, F. and an effect on the performance of the refiner, TAPPI Engineering
Söderberg, D. (2011): Evaluation of steerable filer for detecting Conf., San Francisco, USA.
fibers in flowing suspensions, Exp. in Fluids, Vol. 51, 987 – 996. Lundell, F., Söderberg, L.D. and Alfredsson, P.H. (2011):
Chaussy, D., Martin, C. and Roux, J.-C. (2011): Rheological Fluid mechanics of papermaking, Annu. Rev. Fluid Mech., Vol.
behavior of cellulose fibre suspension: application to paper- 43, 195 – 217.
making processing, Ind. Eng. Chem. Res., Vol. 50, 3524 – Mohlin, U.-B. (2010): Inside an LC-refiner, 11th Biennial
3533. Advances in refining and mechanical pulping: scientific
Derakhshandeh, B., Kerekes, R.J., Hatzikiriakos, S.G. and technical advances, Stockholm, Sweden, Feb 23 – 25.
Bennington, C.P.J. (2011): Rheology of pulp fibre Muhic, D., Huhtanen, J.-P., Sundström, L., Sandberg, C.,
suspensions: a critical review, Chem. Engin. Sci., Vol. 66 (15), Ullmar, M., Vuorio, P. and Engstrand, P. (2011): Energy
3460 – 3470. efficiency in double disc refining – influence on intensity by
Fox, T.S., Brodkey, R.S. and Nissan, A.H. (1981): Inside a segment design, Nordic Pulp and Paper Res. J., Vol. 26 (3),
disc refiner, TAPPI Papermakers Conf., Chicago, USA, April 6 - 224 – 230.
8, 135 – 150. Peskin, C.S. (2002): The immersed boundary method, Acta
Fox, T.S., Brodkey, R.S. and Nissan, A.H. (1982): Inside a Numerica, 479 – 517.
disc refiner, TAPPI J., Vol. 65 (7), 80 – 83. Petrie, C.J.S. (1999): The rheology of fibre suspensions, J. of
Halme, M. (1962): Visual observation on flow phenomena in a non-Newtonian Fluid Mech., Vol. 87 (2-3), 369 – 402.
conical refiner, Paper ja Puu, Vol. 44 (12), 658 – 660. Radoslavova, D., Silvy, J.and Roux, J.-C. (1996): The
Hammarström, D. (2004): A model for simulation of fibre concept of apparent viscosity of pulp for beating analysis and
suspension flows, Licentiate Thesis, Royal Institute of development of the paper properties, TAPPI Papermakers
Technology (KTH), Stockholm, Sweden. Conf., Philadelphia, USA, 195 – 206.
Hietanen, S. and Ebeling, K. (1990): Fundamental aspects of Roux, J.C. (2001): Review paper on pulp treatment process,
the refining process, Paper ja Puu – Paper and Timber, Vol. 72 The Science of Papermaking, Trans. 12th Fund. Res. Symp,
(2), 158 – 179. Oxford, 19 – 80.
Hemström, G., Moller, K. and Norman, B. (1976): Boundary Ryti, N. and Arjas, B. (1968): Influence of residence time
layer studies in pulp suspension flow, TAPPI J., Vol. 59 (8), 115 distribution of the flowing stock on the beating effect of a
– 118. beating machine, Paperi ja Puu, Vol. 50 (11), 661 – 668.
Herbert, W. and Marsh, P.G. (1968): Mechanics and fluid Sjöström, K. (1993): Effect of fibre floc dimensions on the
dynamics of a disc refiner, TAPPI J., Vol. 51 (5), 235 – 239. beating results, Tappi Papermakers' Conf., Atlanta, USA, Book
1, 275 – 289.
Heymer, J.O., Olson, J.A. and Kerekes, R.J. (2011): The role
of multiple loading cycles on pulp in refiners, Nordic Pulp and Tornberg, A.-K. and Shelley, M.J. (2004): Simulating the
Paper Res. J., Vol. 26 (3), 283 – 287. dynamics and interactions of flexible fibres in Stokes flows, J.
Comp. Phys., VOl. 196 (1), 8 – 40.
Huhtanen, J.-P., (2004): Modeling of fibre suspension flows in
refiner and other papermaking processes by combining non- Wu, J. and Aidun, C.K. (2010): A method for direct simulation
Newtonian fluid dynamics and turbulence, PhD Thesis, of flexible fiber suspensions using lattice Boltzmann equation
Tampere University of Technology, Tampere, Finland. with external boundary force, Int. J. Multiphase Flow, Vol. 36,
202 – 209.
Hämäläinen, J., Lindström, S.B., Hämäläinen, T. and
Niskanen, H. (2011): Papermaking fibre-suspension flow Wikström, T. (2002): Flow and rheology of pulp suspensions at
simulations at multiple scales, J. Eng. Math., Vol. 71 (1), 55 – medium consistency. Doctoral Thesis, Dept. of Chemical
79. Engineering, Chalmers University of Technology, Göteborg,
Sweden.
Kondora, G. and Asendrych, D. (2009): Flow modelling in a
disc refiner, 14th Int. Conf. on Fluid Flow Technologies, Yakhot, V., Orszag, S.A., Thangam, S., Gatski, T.B. and
Budapest, Hungary, Sept. 9 – 12. Speziale, C.G. (1992): Development of turbulence models for
shear flows by a double expansion technique, Physics of Fluids
Krochak, P.J., Olson, J.A. and Martinez, D.M. (2009): Fibre A, Vol. 4 (7), 1510 – 1520.
suspension flow in a tapered channel: The effect of flow/fibre
coupling, Int. J. Multiphase Flow, Vol. 35 (7), 676 – 688.
Lee, P.F. and Duffy, G.C. (1976): Velocity profiles in the drag
reducing regime of pulp suspension flows, Appita, Vol. 30 (3),
219 – 226.
Lindstedt, M., Huhtanen, J.-P. and Karvinen, R. (2009):
Optimization of pulp refiner performance, Papermaking
Research Symposium, Kuopio, Finland, June 1 – 4.

Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 183