investigation of sample-size effects on in-plane … of sample-size effects on in-plane tensile ......
TRANSCRIPT
Investigation of sample-size effects on in-plane tensile testing of paperboard Anton Hagman and Mikael Nygårds
KEYWORDS: Size dependency, Paperboard, In-plane,
Tensile test, Speckle photography
SUMMARY: The impact of sample size on in-plane
strain behavior in paperboard was investigated, with the
aim to explore the differences between local and global
properties in paperboard, and try to pinpoint the
mechanisms behind such differences. The local properties
are of interest in converting as well as for future 3D
forming of paperboard. It is important to identify
differences in behavior between local and global
properties since most paperboards are evaluated against
the latter. The methods used for evaluation were tensile
tests in controlled environment and speckle photography.
The results show that there is a difference in strain
behavior that is dependent of the length to width ratio of
the sample, that this behavior cannot be predicted by
standard tensile tests and that it depends on the board
composition. The speckle analysis revealed that the
behavior is a result of the activation of strain zones in the
sample. These zones are relatively constant in size and
therefore contribute differently to total strain in samples
of different size.
ADDRESSES OF THE AUTHORS: Anton Hagman
([email protected]), Mikael Nygårds ([email protected]),
KTH, BiMaC Innovation, Hållfasthetslära, Osquars backe
1, 10044 Stockholm Sweden
Corresponding author: Anton Hagman
In testing of papers large test specimens are often used in
standards. Hence, global properties that have a tendency
of measuring the weakest point in the paper are
characterized. These ''global'' properties may vary from
local properties due to papers heterogeneous nature. The
global properties represent, at least in tension, the
weakest local properties for strength and the average
properties for local strain. Therefore they often suffice to
predict the quality of the paper. But in some cases, e.g.
during converting, the behavior diverge from what is
predicted by the global properties, since the operations
carried out only affect a small part of the board being
converted. An example is a fold that affects a short but
wide area of the paper, compared to the standard
sample’s long and slender geometry.
When it comes to failure in tension it is often caused by
local deformation in a weak spot, such as a defect or an
area with lower density. Such local deformations are
often occurring at fiber-length scale or smaller.
If such local behavior can be identified at an early stage
it can be used or avoided. This would be of interest for
converting or to improve 3D forming methods.
There are indications that denser parts or flocks in paper
can indirectly cause failure, since they are stiffer and
therefore force nearby weaker parts of the paper to higher
local strains (Niskanen 1998, Norman 1965).
Experiments on paper with different strip lengths,
performed by Malmberg (1964), showed that an increase
in strip length cause a decrease in strip strength and
stretchability (lower stress and strain at break),
furthermore the Young's modulus increased. As an
example the strain at break in the cross direction
decreased from 8.73 % to 7.09 % when the strip length
was changed from 50 mm to 200 mm for copy paper.
These experiments were conducted on newspaper, copy
paper, and kraft liner. The decrease in stress at break was
attributed to weakest link behavior by Niskanen (1998).
Furthermore Tryding (1996) briefly mentions a relation
between strain and the sample width to length ratio in
paperboard. A change of width to length ratio from ~0.3
to 3 resulted in a change in strain at failure from 0.025 to
0.055 for a 120 g/m2 paperboard.
Some experiments using different imaging techniques to
collect strain data on paper has been conducted, examples
are; Lyne and Bjelkhagen (1981) who use laser
spectrometry to study the effects of clamping, Choi et al.
(1991) who show that photo spectrometry can be used on
paper and, Weins et al. (1998) who focus on the strain
behavior around the notches in a double edge notched
tension (DENT) sample to investigate fracture behaviors
of paper. All these studies were carried out on copy
paper.
A more recent study was performed by Considine et al.
(2005) who made a good overview of previous literature
on local strain field behavior, before setting out to
characterize the strain fields using speckle photography.
They concluded that speckle is an effective way to
examine local strains in fibrous networks. In their work
they found that there is a transition point between
homogeneous and heterogeneous strain behavior
although they didn’t correlate this transition to any
special event.
Korteoja et al. (1996) made a study, also referenced by
Considine, in which they detected micro damage in
silicone treated papers. The damage was detected as lines
in the strained paper. The damage occurred at less than
half of the failing strain, when the stress strain curve
became non linear. The lines propagated in diagonal
directions. There were few lines detected in samples
strained in the machine direction. In a later study
Korteoja et al. (1998) correlated the microscopic damage
behavior from the previous study to local strains using
speckle. Ostoja-Starzewski and Castro (2003) cross-
correlated the formation and strain field using finit
element simulations, concluding that the higher the basis
weight, the higher local stiffness and strength of the
paper.
The aim of this study was to examine how the sample
geometry affects the tensile properties of multiply
paperboard using speckle photography and regular tensile
tests. Apart from rectangular samples, notched samples
PAPER PHYSICS
Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 295
were also studied in an attempt to even further localize
the properties. The advantage with such a small sample,
i.e. the nick, is that it is held by its natural surroundings,
i.e. not tightly clamped. The disadvantage is the problem
of distinguishing between effects in the nick and effects
caused by the surrounding board.
Materials and Methods The experiments were conducted on three different
multiply paperboards. Two of them are 5 ply boards
(board B and C), from the same mill, made to match the
same specifications but with a different fiber composition
in the different layers (and slightly different grammage).
The third board (board A) is a 3 ply board from another
mill. Descriptions of the different boards can be found in
Table 1.
The boards were tested in four different sets of samples.
Set 1 and 2 had constant length (10 mm resp. 25 mm)
with varying width (50, 25, 15, 10, 5 and 2.5 mm). Set 3
had constant width (15 mm) and varying length (100, 50,
25 and 10 mm). Set 4 had notched samples with a notch
length (Ln) of 0.9 mm and a varying nick width (Wn) of 5,
2.5, 0.8 and 0.2 mm, as shown in Table 2. Table 2 also
contains the measurements for Set 5, in which board B
was tested at different strain rates. A sketch of a notched
sample can be seen in Fig 1.
The notched samples had a clamping length Lc of 40
mm and a width Ww of 15 mm. All samples had
approximately 10 mm excess material that was clamped.
All five sets were conducted in a controlled environment
(23°C and 50% RH) using a tensile test machine (MTS)
with displacement controlled piston movement (speed
0.25 mm/s for set 2, 3, 4 and 0.4 mm/s for set 1). A
complementary set, Set 5, was made with one of the
boards to determine the effect of the difference in relative
strain rate. In this set three different lengths (100, 25 and,
10 mm) was tested at three different strain rates (25 % s-1
,
1-4 % s-1
, and 0.25 % s-1
). The data for the middle set was
taken from earlier sets, and therefore varied a bit (1 % s-1
for the 25 mm long samples, and 4 % s-1
for the 10 mm
long samples). For Set 1, 2 and 5, a 1 kN load cell was
used, for Set 4 a more sensitive 500 N load cell was used.
In Set 1 and 2 four samples of each size were tested. For
the notched samples in Set 4, six samples per size were
used. All sizes were tested in both machine direction
(MD) and cross direction (CD). In addition to this Set 4
was also tested in the 45°-direction.
The samples in Set 1, 2, 3 and 5 where cut using
standard paper cutters (guillotine and rotary). The
samples in Set 4 were first cut into 15 mm wide strips of
appropriate length. Nicks were produced in the strips by
grinding two opposite 0.9 mm wide notches in the middle
of the strip. The notches were ground using a rotating
diamond plated cutting wheel. The strips were ground
five or six at a time, with two sacrificial strips on either
side. The samples were made with four different target
nick widths Wn= 0.2, 0.8, 2.5 and 5.0 mm. The final
width differed a bit between individual samples.
A batch of notched samples (Wn = 5 mm) with a printed
speckle pattern was also made. The pattern was printed
on the paperboard in a standard laser printer before the
samples was ground in the same way as above. The
samples where then loaded in tension in a tensile testing
machine (Instron), while being recorded by speckle
equipment (Aramis). These tests were performed in the
MD and CD directions and outside controlled climate.
Furthermore unnotched specimens of different sizes were
tested and photographed.
All samples tested in controlled climate were carefully
investigated such that no slipping had occurred in the
clamps. However, the clamps used for the speckle tests
were smooth and the risk of slippage was therefore
higher. Nevertheless any slipping could easily be
observed from the stress-strain curve which would then
form a “saw-tooth” pattern, and thus be controlled for.
Strain mapping was displayed using both a continuous
and a discretized scale, the latter made it possible to
estimate zone sizes compared to the sample size. This
was done by examining the percentage of the different
discreet strain levels, using image analysis in Matlab.
Furthermore the strain profile, along straight lines parallel
to the samples, was examined. The lines chosen were
representative for the whole sample, i.e. they had the
same development as the whole sample with regard to
mean, minimum, maximum and deviation of the strain.
The mentioned quantities for both lines and the sample as
a whole, was calculated directly in the Aramis software.
Table 1 Properties of the different boards used.
Board # ply Thickness [mm]
Grammage [g/m2]
Density [kg/m3]
A 3 0.31 210 677
B 5 0.35 220 628
C 5 0.33 230 657
Table 2 Sample dimensions for the different sets.
Set Width Ww [mm]
Length Lc [mm]
Nick- width Wn [mm]
1 50, 25, 15, 10, 5, 2.5 10 -
2 50, 25, 15, 10, 5, 2.5 25
3 15 100, 50, 25, 15, 10, 5
-
4 15 40 5, 2.5, 0.8, 0.2
5 15 100, 25, 10 -
Fig 1. Sketch of notched samples with the measurements used in Table 2 defined. F shows the direction of applied force during testing. Ln is the length of the notch.
PAPER PHYSICS
296 Nordic Pulp and Paper Research Journal Vol 27 no.2/2012
Fig 2. Stress strain curves for board B - set 1. Constant length (10 mm) with varying widths. Dashed lines indicate CD samples and solid lines MD samples. In this graph a representative sample for each length has been selected for clarity.
Fig 4. Stress strain curves for board B - set 2. Constant length (25 mm) varying widths. Dashed lines indicate CD samples and solid lines MD samples. In this graph a representative sample for each length has been selected for clarity.
Results and discussion Tensile tests in controlled climate
Using the force-displacement curves, stress-strain curves
were calculated using the length, width and thickness of
the samples. These calculations were performed for Set 1,
2, 3 and 5. For Set 4, only the stress was calculated. An
attempt to calculate the strain was made, but showed too
small correspondence to the strains measured with
speckle results to be deemed useful. Fig 2-Fig 4 show typical stress-strain curves for board
B Set 1, 2 and 3. Typical stress-displacement curves for
notched samples of board B (Set 4) can be found in Fig 5.
From the stress-strain curves, for Sets 1, 2 and 3 for
board B, it seems that the yield point and the plastic
hardening varied for different sample sizes. The yield
point seems to correlate to the hardening. This behavior
also correlates to the strain at break, i.e. early yield,
decreased hardening and increasing strain at break goes
hand in hand. From this it might be inferred that the same
mechanism that increases the strain at break also plays a
role for the yield point and the hardening behavior. If this
mechanism could be identified, it could be possible to
predict final strainability of a sample already at, or right
after yielding. Such a mechanism is proposed further
Fig 3. Stress strain curves for board B - set 3. Constant width (15 mm) varying lengths. Dashed lines indicate CD samples and solid lines MD samples. In this graph a representative sample for each length has been selected for clarity.
Fig 5. Stress strain graph for board B - set 4. Notched samples (Nw: 5-blue, 2.5-green, and 0.8-black, 0.2-red). Dashed lines indicate CD samples and whole lines MD samples. In this graph a representative sample for each length has been selected for clarity.
down based on the speckle analyzes. From the curves it is
still apparent that the main difference in strainability
between the different sample sizes occurred in the
“plastic” regime. This is consistent with observations
made by Korteoja et al. (1996).
All three board types display the same kind of behavior
for sets 1, 2 and 3. That is, a fairly constant stress at break
for all length to width ratios. A drastic increase in strain
at break occurs when the length to width ratio decreases
below 1. The different sets have the same behavior not
only qualitatively but also quantitatively (i.e. the curves
do not only look the same, they also have roughly the
same magnitude). The absolute strain increase is similar
in both CD and MD, causing a larger relative strain
increase for MD samples. This can be seen in Fig 6,
where the strain at break has been plotted as function of
length/width ratio for board B. In Fig 7 and Fig 9 the
corresponding behaviors for boards A and C can be seen.
The change in strain for Set 1 and 2 shows that wider
samples result in an increasing strain at break. This
phenomenon will be revisited and explained further
down, using reasoning based on the speckle analysis.
PAPER PHYSICS
Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 297
Fig 6. Strain at break plotted against length/width ratio for board B. Red markings notched sample (Wn= 5 mm).
Fig 7. Strain at break plotted against length/width ratio for board C. Red markings notched sample (Wn = 5 mm).
Fig 8. Boards from Set 1 compared to each other. The different boards show the same behavior in MD. In CD the strain for board C is much more influenced by the sample size then the other boards.
Fig 9. Strain at break plotted vs length/width ratio for board A.
Fig 10. The effect of strainrate on strain at break for board B, lengths 10, 25 and, 100 mm. Rates 0.25, ~2 and, 25 % s-1.
Fig 11. Boards from Set 2 compared to each other. All boards show the same behavior in both CD and MD.
PAPER PHYSICS
298 Nordic Pulp and Paper Research Journal Vol 27 no.2/2012
Fig 12. Boards from Set 3 compared to each other. The different boards show the same behavior in MD. In CD the strain for board C is much more influenced by the sample size then the other boards.
The results from Set 5, as seen in Fig 10, shows that the
impact of the difference in relative displacement speed on
strain at break was small compared to the difference due
to the length of the sample.
In Figs 8, 11 and 12 comparisons between the different
boards for the different sets are made. The behavior in
MD was almost identical for board B and C, while board
A had the same sort of behavior but with an overall lower
strain. In CD, all boards show the same trends, but board
A had an equal or higher strain than board B. Both other
boards were superseded by board C which had a more
rapid increase in strain as the length to width ratio
decreased.
From the comparison between boards it was obvious
that boards that seem to be equal in a standard test have
quite different local properties (as seen for Set 3 in Fig 12
or Set 1 in Fig 8. This indicates that changes made to
improve standard test results might have a negative effect
for the local properties and thus have an unintended
effect on converting or forming of the board. As stated
above this effect was exclusively seen in CD. The reason
for this can partly be found in the speckle pictures where
the strain distribution was much more homogeneous in
MD than in CD for short wide samples. Another
interesting aspect of these graphs was the difference
between the boards. A comparison between board B and
C shows that the denser board has “better” local
properties.
Speckle analysis
In Fig 13 and Fig 14 speckle images of samples with size
100 x 50 mm2 in MD respectively CD can be seen. The
measured area on the samples was smaller than the whole
sample (~80x45 mm2) and represents the strain state prior
to break.
Strain zones or streaks can be seen cross the samples. In
these zones the strain was much larger than in the rest of
the sample. The zones were much more prominent in
samples pulled in CD, where they form approximately 5
mm wide streaks corresponding to a few fiber lengths
(compare to the grids in figures Fig 13 and Fig 14 which
have 10x10 mm2 cells). These streaks are party wise
diagonal across the sample, this is once again consistent
with previous observations (Korteoja et al. 1996). In the
MD tested samples, the streaks are more diffuse, but
zones clearly appear. The streaks became noticeable, in
both directions, when the sample started to show a
“plastic” behavior. For the shortest samples the streaks
were most diffuse. The perpendicular strain was also
largest in areas corresponding to the strain streaks seen in
the straining direction.
The exceptions were some of the short wide samples,
e.g. 25x50 mm2 and 10x50 mm
2, where a concave strain
pattern was observed. This pattern only occurred in
samples strained in MD. It was suspected to be the result
of pull of slightly skew fibers that are clamped in one
end, with the other end pulling in the edge areas where
there were shorter fibers and perhaps more loosened
bonds. The effect would be cancelled out towards the
middle of the sample.
Fig 13. Strain distribution in part of a 100x50 mm2 MD sample just prior to break. Each grid square is 10 by 10 mm.
PAPER PHYSICS
Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 299
Fig 14 Strain distribution in part of a 100x50 mm2 CD sample just prior to break. Each grid square is 10 by 10 mm.
The strain streaks can explain the increase in strain at
break for shorter wider samples. As the samples become
shorter the streaks occupy a larger portion of the sample.
Fig 15 and Fig 16 show a discretized strain distribution
for the different samples sizes prior to break. Fig 17 and
Fig 18 show how much of the sample that is covered by
strain zones, for MD and CD respectively. In these
graphs an area with a local strain over 2.1 resp. 4.2 % is
considered to be a strain zone (i.e. the parts of Fig 15 and
Fig 16 that is yellow or red). For the CD samples it is
clear that shorter samples have a larger amount of strain
zones than long samples of the same width. It is also clear
that wider sample has a larger part of the total area
strained than narrow samples of the same length. In the
narrower samples there is just one such zone, and in the
wider samples there are more than one. The zones are of
roughly the same size in all samples, thus occupying a
relatively larger part of the sample per zone for smaller
samples. The MD samples behave differently, the relative
zone area increases with shorter samples, but it is only for
the longest samples that the wider sample has an
increased zone area. For the 10 and 25 mm long samples
the relative zone area decreases for wider samples.
Fig 15. MD samples of different sizes with discretised strain distributions. From top left: 10x50, 10x15, 25x50, 25x15, 100x50 and 100x15 mm2. Note that the size scale is only approximate between the samples.
A possible reason for this is that the perpendicular strain
behaves differently in these samples, as mentioned above.
While a wide sample increases the possibility for a
starting point for such a zone, it also decreases the
possibility for that zone to immediately cross the whole
sample, making it possible for other parts of the sample
to elongate.
One arguable explanation for these zones can be derived
from the arguments of Norman (1965); due to the
inhomogeneity of the board, there are stronger and
weaker parts. The stronger parts force the weaker parts to
deform. Korteoja et al (1996) related damage streaks to
fibers oriented transverse to the straining direction. They
also point out that the site of failure often can be spotted
before the actual break.
One possible explanation is that interfiber bonds in the
weaker parts are broken and the fibers start to align in the
straining direction. This further weakens the zone
compared to the rest of the sample. As mentioned above
the strain streaks start to form when the stress strain curve
becomes nonlinear. This is also consistent with Korteojas
findings. An example of this can be found in Fig 19
where the strain distribution in the beginning of the
nonlinear region is displayed, for the same 100x50 CD
PAPER PHYSICS
300 Nordic Pulp and Paper Research Journal Vol 27 no.2/2012
Fig 16. CD samples of different sizes with discretised strain distributions. From top left: 10x50, 10x15, 25x50, 25x15, 100x50 and 100x15 mm2. Note that the size scale is only approximate between the samples.
sample as in Fig 14. A visual comparison between the
samples shows great similarity of the zones in the
different time steps (note the different scales). In Fig 21
the strain profile along the line indicated in Fig 19 is
displayed. The strain profiles at different time steps are
displayed along with the same lines normalized against
the mean strain along the line in each step. Finally a
somewhat simplistic comparison is made between the
strain profiles along the line at different time steps. The
graph simply shows how the peaks and valleys of the line
are positioned compared to the profile just before
breaking. This is done by seeing which parts of the line
that is respectively above and below the mean strain of
the line. The blue lines in the graphs are for profiles
captured in the non linear part of the stress strain curve
and the red line is in the linear part. It is apparent from
the graphs and Fig 19 that the zones form at an early
stage, which is consistent with beginning/increasing
inter-fiber bond breakage. The same behavior was
observed for the other sample sizes in CD and a
somewhat lesser degree in MD.
Fig 17. Relative area of samples that is covered by strain zones for different CD samples. Parts of sample with local strain above 4.2% are considered to be strain zones.
Fig 18. Relative area of samples that is covered by strain zones for different MD samples. Parts of sample with local strain above 2.1% are considered to be strain zones.
The early occurrence of the strain zones suggests that it
is the zones that are responsible for the differences in
hardening and yield behavior between different sample
sizes. The zones appear to have a different modulus than
the rest of the sample, at least if the stress in the sample is
considered to be uniform. Simulations made by Korteoja
et al. (1996) support this interpretation. If the zones have
a different modulus than the rest of the sample, it is not
farfetched to reason that they also have a different
hardening behavior. This is supported by the behavior of
the standard deviation of the strain, shown in Fig 20 and
Fig 22, as the mean strain increases the variation in the
sample increases. This increase is fairly linear in the CD
samples and more exponential in the MD samples. These
differences would change the overall behavior of the
sample based on the relative area of the zones. This
would explain the differences in the stress strain curves,
mentioned earlier. The possibility for the zones to be
detected with high accuracy at an early stage in the non
linear region (as suggested by the graphs), should be
interesting from an engineering viewpoint. If the zones
can be detected by some other technique that does not
PAPER PHYSICS
Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 301
Fig 19. The strain distribution in a 100x50 mm2 CD sample (same Fig 16) in the beginning of the non linear part of the stress strain curve. The blue line indicates where the strain profiles in Fig 21 are taken. The Z+ seen in the top of the picture is indicating the viewing direction in the Aramis-software.
Fig 20. Standard deviation of the strain in all measuring points on the whole sample plotted against the mean strain for the whole sample (as measured by speckle) for each measuring step. CD samples.
require a printed pattern or other destructive interference
with the paper sheets (i.e. a thermo-camera), then maybe
the weak zones can be strengthened or the strong zones
weakened, and in such a way increase the overall strain
performance of the sheet.
No apparent damage can be seen on the surface of the
sample until just prior to break, which might infer that the
damage zone process is mainly happening in the middle
layer. In the paperboards the top ply contains a larger
amount of softwood fines, than the bulky middle plies. It
can therefore sustain a higher strain than the bulk layer.
This would explain why converted paperboard can
sustain the high strains on the outside of a fold without
Fig 21. Top: Strain profiles along the line indicated in Fig 19, at different times during straining. Top line is just prior to break, the red line is taken when the sample is still in the linear region. Middle (top): Strain profiles normalized with mean strain for the whole line. Middle (bottom): Stress strain curve with sample times marked. Bottom: Position of peaks and valleys at different times compared to top line.
Fig 22. Standard deviation of the strain in all measuring points on the whole sample plotted against the mean strain for the whole sample (as measured by speckle) for each measuring step. MD samples.
apparent damage. A question that arises is if the final
failure was due to a drop in load bearing in the middle
layer, or because of increased damage in the top and or
bottom layers.
Notched samples
The stress was plotted against total sample displacement
instead of strain for Set 4. This was done since the
speckle samples showed that the strain situation in the
samples was rather complicated (see Figs 23 and 24).
PAPER PHYSICS
302 Nordic Pulp and Paper Research Journal Vol 27 no.2/2012
Fig 23. Speckle image for notched CD sample of board B. The strain distribution is superimposed over the original camera picture.
Fig 24. Speckle image for notched MD sample of board B. The strain distribution is superimposed over the original camera picture.
What can be concluded is that the stress at break was
quite constant for the samples and that the total
displacement for the samples decreased with decreasing
notch width. A strain measure can be estimated from the
speckle results, and this is close to the maximum strain of
the samples with width to length ratio > 1.
Samples oriented 45° from the machine direction were
also tested. The results from those tests show that these
samples behave more like the CD samples when it comes
to stress, and more like the MD samples with regard to
displacement, i.e. it follows the direction with lower
values.
Furthermore, wide nicks showed a more brittle
behavior, i.e. the stress-displacement curve dives abruptly
without a trailing tail, for all directions.
The speckle analysis of samples with 5 mm wide
notches shows that the strain was not contained within
the notch; it spreads into the full width parts of the
sample. In CD the strain seems to be focused quite close
to the notch, see Fig 23. In MD the strain spreads deeper
into the full width part, see Fig 24. The difference in
behavior can be attributed to the difference in fiber
direction. The fact that the board was stretched in other
parts of the sample rather than at the notch, explains how
a large total displacement can occur without extreme
strains. It is also worth noting that the strain zone had a
Fig 25. Close up on notched MD sample during crack-propagation. Board B. Cracks can be seen propagating from both sides of the nick.
Fig 26. Nicked MD samples after testing, board B. From left to right: 5, 2.5 ,0.8 and 0.2 mm nickwidth.
Fig 27. Nicked CD samples after testing, board B. From left to right: 5, 2.5 ,0.8 and 0.2 mm nickwidth.
PAPER PHYSICS
Nordic Pulp and Paper Research Journal Vol 27 no.2/2012 303
triangular and/or half circle shape in both directions. This
zone is more triangular in MD while being more circular
in CD. This indicates that there might be a relationship
between notch width and zone length, which would
explain why less wide notches can withstand smaller
displacements. The less brittle break of the less wide
samples further supports this, since a shorter strained
zone outside the sample was unable to store the same
amount of elastic energy, which in turn would make the
break less abrupt.
No damage zones or streaks can be seen in or around
the nicks. This is logical since the nicks are smaller than
the zones and therefore relatively homogeneous. When
the stress in the sample was increased to a point where
bonds began to break, the whole sample broke apart. This
was supported by the tensile curve which had a smoother
transition from elastic to plastic part in the nicked
samples. A possible reason for this is that the “plasticity”
in nicked samples was only due to movements, e.g.
straitening, of free fiber segments that can occur without
the breaking of bonds, while the curve for larger samples
gets a sharper transition due to the activation of the
damage behavior. The lack of zones also explains why
the yield and hardening behavior of the stress strain
curves differs from the behavior seen for unnotched
samples. Once again it should be noted that the speckle
analysis only shows the strains in the top layer of the
board.
The cracks that appeared in the nicks often started in the
corners of the notch, which was not surprising since
stress concentrations were expected there. The crack-path
then varies, often going partly diagonally, and sometimes
propagating out of the nick (the crack always starts and
ends in the nick). For both CD and MD samples the crack
path becomes straighter with narrower nicks. As can be
seen in Fig 25 the cracks in the notched samples can start
on both sides simultaneously, and drives a small strain
field in front of them. Worth noting was that the cracks
and the strain field was smaller than the nick length,
which made it possible to dismiss that the notches
behaved as crack tips. The cracks in the nicks often
propagated from the corners of the notch where stress
concentrations where to be expected.
Fig 26 and Fig 27 show the separated cracks of tested
nicked samples. In these pictures pulled out fibers can be
seen in both MD and CD samples. This indicates that the
whole nick had been activated during the stretching and
that fiber bonds rather than fibers have been broken.
Conclusions There was an effect on the strainability caused by the size
of the tested sample. This effect was related to the length
to width ratio of the sample, and was caused by strain
streaks in the sample. The strain of each individual streak
is fairly constant which means that the number of streaks
and their part of the whole length determines the final
strain at break of a sample. A wider sample reduces the
risk of streaks crossing the sample and thus increases the
overall maximum strain of the sample. These strain
streaks were caused by a damage behavior in weaker
spots in the board and needed a certain amount of space
to be activated. Apart from the strain at break the weak
streaks affect the yield and hardening behavior of the
sample. The zones are detectable at an early stage using
speckle analysis.
Among the board qualities that were tested, the strain
ability of the top layer seems to be significantly higher
than that of the whole sample as such. Furthermore
denser boards showed better local properties i.e. higher
maximal strains.
Acknowledgements
The authors would like to thank the department of solid mechanics at KTH and BiMaC Innovation for financial support. Furthermore the authors would like to thank Innventia for access to their climate laboratory. A special thanks to Sune Karlsson (Innventia) and Veronica Wåtz (KTH, Solid Mechanics) for sharing their expertise.
Literature
Choi, D., Thorpe, L.J, Hanna, R.B. (1991): Image analysis of measure strain in wood and paper, Wood Sci. Technol., 25, pp. 251-262.
Considine, J.M., Scott C.T., Gleisner R., Zhu J.Y. (2005): Use of digital image correlation to study the local deformation field of paper and paperboard, 13th Fund. Res. Symp. Cambridge, 2005.
Korteoja, M.J., Lukkarinen, A., Kaski, K., Gunderson, D.E., Dahlke, J.L, Niskanen, K. (1996): Local strain fields in paper, Tappi J, 79(4), 217.
Korteoja, M.J., Niskanen, K., Kortschot M.T., Kaski, K.K., (1998): Paperi Puu, 80(5),364.
Lyne, M.B., Bjelkhagen, H. (1981): The Application of Speckle Inteferometry to the Analysis of Elongation in Paper and Polymer Sheets, Pulp Paper Can., 82(6), pp. TR29-TR35.
Malmberg, B. (1964): Remslängd och töjningshastighet vid spänningtöjningsmätningar på papper, Svensk Papperst., 67(17), pp. 690-692.
Niskanen, K. (1998): Paper physics, Fapet, Helsinki, pp. 140,171-173.
Norman, R.J. (1965): Dependence of sheet properties on formation and forming variables, Consolidation of the paper web. Trans. IIIrd Fund. Res. Symp. Cambridge, 1965, pp. 269-298.
Ostoja-Starzewski, M., Castro, J. (2003): Random formation, inelastic response and scale effects in paper, Phil. Trans. R. Soc. Lond. A, 361, pp. 965-985.
Tryding, J. (1996): In-Plane fracture of paper. Ph. D. Thesis Lunds Universitet, Lund, Sweden.
Wiens, M., Göttsching, L., Dalpke, B. (1998): Quantifizierung lokaler Verformungen von Papier mit Hilfe einer neuen Meßtechnik, Das Papier, 11, pp. 649 - 654
PAPER PHYSICS
304 Nordic Pulp and Paper Research Journal Vol 27 no.2/2012