caracterización reológica del queso inglés
TRANSCRIPT
-
7/28/2019 Caracterizacin reolgica del queso ingls
1/5
3rd International Symposium on Food Rheology and Structure
267
RHEOLOGICAL CHARACTERISATION OF CHEESE
SM Goh1, MN Charalambides
1, S Chakrabarti
2, JG Williams
1
1Department of Mechanical Engineering, Imperial College London, SW7 2BX, U.K.
2General Mills Technology, Minneapolis, MN 55414
ABSTRACT
A scheme has been developed to characterise the strain
and the time dependent behaviour of non-linear
viscoelastic materials such as cheese in the form of a
non-linear constitutive model. The model consists of two
independent functions - a hyperelastic function to
characterise the strain dependent behaviour, and a Prony
series to characterise the time dependent behaviour. The
calibration of the model is made using data obtained from
monotonic uniaxial compression and stress relaxation
tests. In order to verify the material parameters obtained
from this scheme, experimental tests and finite element
simulations of the three point bend and wire cutting tests
of two cheeses were performed. The results from the
finite element simulations showed good agreement with
the experimental test data under various test conditions.
1. INTRODUCTION
Many foods such as cheese and dough exhibit large
strain, viscoelastic behaviour. For these foods, boththe strain and the time dependent mechanical
behaviour must be characterised. In order that the
constitutive models have good predictive
capabilities, they have to be calibrated using
consistent material data. However, for foods such
as cheese, consistent material data is not obtainable
because of the significant material variation between
different blocks and batches (Prentice et al. 1993).
Thus, the material data is not accumulative and has
to be collected for each batch. Successful methods
for characterising these foods would have to besimple, quick and be economical in terms of time
and materials. In this study, a method for
characterising the non-linear viscoelastic properties
of cheese was investigated. These properties were
then used in analysing three point bend and wire
cutting tests.
2. EXPERIMENTS
Mild Cheddar and Gruyere samples were bought
from a local supermarket and stored at 4C until
testing. A separate block of each cheese was usedfor the three point bend and the wire cutting tests.
In addition, a block of process cheese was supplied
by General Mills and was used in the wire cutting
tests.
The specimens were cut into rectangular shapes
using a wire cutter and into cylinders using a borer.
They were then wrapped in cling film and allowed to
equilibrate at room temperature (21C) for at least
two hours. All tests were performed at 21C using
the Instron 5543 testing machine.
Rectangular specimens of height 15mm, width30mm and length 60mm were prepared for the three
point bend test. The striker and the supports
consisted of steel rods of 10mm diameter with the
supports positioned at 50mm apart. For Gruyere,
the tests were conducted at 5, 50 and 500mm/min.
For mild Cheddar, the tests were conducted at two
crosshead speeds, 5 and 50mm/min. A further test
was also performed at 5mm/min using mild Cheddar
specimens which were notched to a depth of 7.5mm
at the plane of symmetry.
The wire cutting tests for Gruyere and mild Cheddarwere performed using wire diameters, d , of 0.25,
0.5 and 0.89mm as well as dowel pins of diameter
1.6 and 2mm. For the process cheese, wire
diameters of 0.25, 0.345, 0.5 and 0.89mm were
used. The dowel pins were sufficiently rigid, so the
crosshead displacement was an accurate measure
of the displacement of the pins. For the smaller wire
diameters, the crosshead displacement had to be
corrected for the deflection of the wire relative to the
crosshead to obtain the actual wire displacement
(Goh 2002).
The specimens for the wire cutting tests were
rectangular blocks of length 25mm, height 20mm
and width 15mm for the smaller wire diameters.
Blocks of length 30mm, height 30mm, and
thicknesses 20mm and 30mm were used for the
1.6mm and 2mm diameters respectively. Three
constant cutting speeds of 5, 50 and 500mm/min
were used.
The material calibration tests involved monotonic
uniaxial compression tests performed at true strain
rates of 0.25, 2.5 and 25/min, and relaxation tests
-
7/28/2019 Caracterizacin reolgica del queso ingls
2/5
3rd International Symposium on Food Rheology and Structure
268
performed at a true strain rate of 2.5/min up to a
strain of 0.04. The specimens were cylinders with
height 20mm and diameter 20mm. Prior to the start
of test, the platens were lubricated with Superlube
(Loctite Corp.) to eliminate the friction at the sample-
platen interface (Charalambides et al. 2001). Since
separate blocks of each cheese were used to studythe three point bend and the wire cutting tests, the
material data, in the form of true stress, , and true
strain, , were collected for each block separately.
The true stress was calculated based on the
assumption that the material was incompressible.
3. NUMERICAL SIMULATIONS
The numerical simulations were performed in the
commercial finite element code ABAQUS. In all
models, four noded, plane strain elements were
used to model the cheese. Because of symmetry,
only half of the specimen was required. The striker
and the support were modelled as rigid surfaces.
The striker was prescribed to move at speeds which
are identical to those in the experiments. The
contact surfaces were assumed to be frictionless
since preliminary results showed that the friction had
a negligible effect on the bending force.
For the wire cutting test, the focus of the finite
element analysis was on the indentation of the wire
into the specimen. This phase precedes the steady-
state cutting phase where the wire makes a cut
through the specimen. Only half of the specimen
was included due to symmetry. The contact
between the wire and the specimen was assumed to
be frictionless since the friction was found to have a
negligible effect on the indentation force.
4. THEORY
In ABAQUS (ABAQUS 1998), the viscoelastic model
consists of two independent components which
represent the strain and the time dependentbehaviour. During a step-strain relaxation test, the
relationship between the stress and the time and
strain can be expressed as,
( ) ( )tgf = (1)
where t is the time, and f and tg are the strain
and the time dependent functions respectively.
The time dependent behaviour in ABAQUS is
defined by the Prony series, which is expressed as,
( )
=
+=
N
i ii
tggtg
1
exp
(2)
where i are time constants and ig are dimension-
less numbers, and,
1
1
=+=
N
i
igg (3)
For non-linear, large deformations, the strain
dependent behaviour is defined by a hyperelastic
strain energy potential.
The Prony series and the hyperelastic potential canbe calibrated from ideal relaxation test data and the
stress-strain relationship corresponding to
instantaneous or long term deformation. However, it
is often not possible to perform experiments under
these ideal conditions, as is the case in this work.
Under non-ideal test conditions, the test data can be
described instead by the convolution integral,
( ) ( )( )
ss
fstgt
t
dd
d
0
=
(4)
The solutions of the convolution integral can befitted to experimental data to obtain the material
constants in f and tg . Although this direct
method is feasible for some hyperelastic functions,
such as the Mooney-Rivlin and polynomial strain
energy functions (e.g. Miller 1999), the convolution
integral can become intractable for other forms of
hyperelastic functions. An alternative procedure to
overcome this problem is to first calibrate the strain
dependent behaviour with a polynomial expression
given by,
( ) DCBAf +++= 234 (5)where A , B , C and D are constants. In
combination with the Prony series, the solutions to
the convolution integral can be obtained and
calibrated through a scheme as proposed in Goh et
al. (2002). After equation (5) has been calibrated, it
represents the stress-strain relationship under
instantaneous (i.e. t=0, tg =1 and f= in
equation (1)), uniaxial compression state, to which
other hyperelastic functions can be approximated.
The fitting of the hyperelastic functions is made byinputting the stress-strain data calculated using
equation (5) into ABAQUS. During the pre-
processing stage, ABAQUS automatically
approximates the input data with the chosen
hyperelastic function. The Van der Waals
hyperelastic potential was used in this work because
it led to a good approximation of the data.
Furthermore, it is also known to provide a more
accurate prediction of the general deformation
modes if the calibration of the material constants is
based only on one test (ABAQUS 1998).
-
7/28/2019 Caracterizacin reolgica del queso ingls
3/5
3rd International Symposium on Food Rheology and Structure
269
5. RESULTS AND DISCUSSION
Typical stress-strain curves of the cheeses used in
the wire cutting tests are shown in Figure 1. The
stress-strain data for the cheeses used in the three
point bend tests also share similar characteristics.
There was in general more scatter in the data for the
process cheese. This was due to the sagging of the
cheese under its own weight during the storage
period which led to a rather inhomogeneous mat-
erial.
0
10
2030
40
50
60
70
80
90
100
0 0.1 0.2 0.3 0.4 0.5 0.6
strain
stress(kPa)
=25/min.
=2.5/min.
=0.25/min.
(a)
0
10
20
30
40
50
60
70
80
0 0.1 0.2 0.3 0.4 0.5 0.6
strain
stress(kPa)
=25/min.
=2.5/min.
=0.25/min.
(b)
0
10
20
30
40
50
0 0.3 0.6 0.9 1.2 1.5 1.8strain
stress(kPa)
=25/min.
=2.5/min.
=0.25/min.
(c)
Figure 1 Stress-strain curves for (a) Gruyere (b) mild
Cheddar (c) process cheese
The results of the calibration of the polynomial and
the Prony series are shown in Table 1. Also
included in Table 1 are the values of the Van der
Waals hyperelastic constants. For the calibration,
the values of i were arbitrarily chosen as 0.1, 1,
10, 100 and 1000 seconds for i equal to one to five
respectively.
A (kPa) B (kPa) C (kPa) D (kPa) (kPa) m a
*mild Cheddar -1100 2040 -1330 560 172 3.11 1.51
*Gruyere -3100 4030 -2040 730 230 2.59 2.02
**mild Cheddar -3250 4050 -1950 610 190 2.54 2.19
**Gruyere -4440 5210 -2380 765 236 2.64 1.98
Processed 6.6 -8.5 -7 61.6 22.8 412 0.103
(a)g1 g2 g3 g4 g5 g
*mild Cheddar 0.312 0.289 0.109 0.101 0.109 0.080
*Gruyere 0.117 0.404 0.128 0.133 0.108 0.110
**mild Cheddar 0.304 0.303 0.114 0.106 0.089 0.084
**Gruyere 0.221 0.333 0.117 0.123 0.097 0.109
Processed 0 0.525 0.233 0.201 0.040 0.000
(b)
Table 1 Material parameters for (a) Strain dependent
function (b) Time dependent function *Three point bend
**Wire cutting/Indentation
The comparison between the experimental force-
displacement curves in the three point bend tests
and finite element predictions is shown in Figure 2.
A good agreement is observed in general.
0
2
4
6
8
10
12
0 2 4 6 8 10displacement (mm)
bendingforce(
N)
experimental
finite element
prediction
500mm/min
50mm/min
5mm/min
(a)
0
1
2
3
4
5
6
0 2 4 6 8 10displacement (mm)
bendingforce(N)
experimental
finite elementprediction
5mm/min
500mm/min
5mm/min
notchedspecimen
(b)
Figure 2 Bending force-displacement data (a) Gruyere (b)
mild Cheddar
-
7/28/2019 Caracterizacin reolgica del queso ingls
4/5
3rd International Symposium on Food Rheology and Structure
270
The numerical predictions of the indentation forces
are also in good agreement with the experimental
data for all wire diameters. The results for three
cases are shown in Figure 3. There was a small
scatter in the data for Gruyere and mild Cheddar so
the average values are shown. As mentioned
earlier, there was considerable scatter in the datafor process cheese and so the raw data are shown.
These results show that the proposed scheme for
obtaining the material constants is successful in
characterising the strain and the time dependent
material behaviour.
0
2
4
6
8
10
12
14
0 0.5 1 1.5 2 2.5wire displacement (mm)
force
(N)
500mm/min
50mm/min
5mm/min
finite element
prediction
(a)
0
0.4
0.8
1.2
1.6
2
0 0.2 0.4 0.6 0.8 1wire displacement (mm)
force(N)
500mm/min
50mm/min
5mm/min
finite elementprediction
(b)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 0.5 1 1.5 2wire displacement (mm)
force(N)
500mm/min
50mm/min
5mm/min
finite elementprediction
(c)
Figure 3 Indentation force-displacement data (a) Gruyere,
d=2mm (b) mild Cheddar, d=0.5mm (c) process cheese,
d=0.25mm
The wire cutting models were further investigated for
their ability to predict the steady-state cutting forces.
For this task, a simple fracture criterion based on a
critical strain was used. This fracture criterion was
adopted, because the fracture strains remained
relatively unchanged for the different strain rates for
the cheeses. Thus, the critical strain, crit , wasassumed to be equal to the fracture strain as
measured in the uniaxial compression test. For mild
Cheddar and Gruyere, the global fracture of the
specimens were observed to occur around the peak
in the curves. Thus, the fracture strains are
approximately 0.5 and 0.45 respectively. For the
process cheese, the specimens underwent a high
degree of compression and when the specimen
fractured, no drop in stress was recorded. From
visual observations, the specimens appeared to
fracture at strains of 1.4-1.6.In the finite element indentation models, the
maximum tensile strain, max,xx , occurs at the line of
symmetry in the direction normal to the movement
of the wire. With increasing indentation, the value of
max,xx increases monotonically. Thus, the changes
in max,xx along the line of symmetry were
monitored such that when crit was reached,
fracture was assumed to occur. The numerically
predicted indentation forces per unit width are
compared with the experimental steady-state cutting
data in Figure 4. Good agreement between thepredicted values and the experimental data is
observed. Thus, the critical strain criterion appears
to be valid for the prediction of the cutting force.
The validity of the indentation models to predict the
steady-state cutting force does require further
research. In the indentation models, it was found
that surface friction had a negligible effect on the
indentation load. However, theoretical considerat-
ions of the steady-state cutting stage (Kamyab et al.
1995) have predicted a large influence of the friction
on the cutting force. Furthermore, it has been
assumed that the fracture strain in tension was
equal to the fracture strain in compression. Since
the deformation of the material ahead of the wire is
highly constrained, it was also assumed that the
fracture strain was independent of the hydrostatic
stress. It will be necessary that other independent
tests such as the plane strain compression and the
tension tests be performed to investigate the
material behaviour in deformation states other than
uniaxial compression. The modelling of the steady-
state cutting stage will also be necessary tounderstand more fully the stress and deformation
states as well as the effect of friction.
-
7/28/2019 Caracterizacin reolgica del queso ingls
5/5
3rd International Symposium on Food Rheology and Structure
271
0
100
200
300
400
500
0 0.5 1 1.5 2wire diameter (mm)
cuttingfo
rce/width(J/m2)
5mm/min
50mm/min
500mm/min
finite elementprediction
(a)
0
100
200
300
400
500
0 0.5 1 1.5 2wire diameter (mm)
cuttingforce/w
idth(J/m2)
5mm/min
50mm/min
500mm/min
finite elementprediction
(b)
0
20
40
60
80
100120
140
160
0 0.2 0.4 0.6 0.8 1wire diameter (mm)
cuttingforce/width(J
/m2)
5mm/min
50mm/min
500mm/min
(c)
Figure 4 Prediction of steady-state cutting force for (a)
Gruyere (b) mild Cheddar (c) process cheese solid linerepresents finite element prediction using crit =1.6;
broken line represents finite element prediction using
crit =1.4
6. CONCLUSIONS
Finite element simulations have been performed to
model the mechanical behaviour of cheese. The
material models were calibrated through an indirect
approach, where the strain dependent behaviour
was first characterised by a polynomial, which was
then fitted with the Van der Waals hyperelastic
function in ABAQUS. The time dependent behaviour
was modelled using Prony series. The numerical
force-displacement curves were in good agreement
with the experimental data for three point bend and
indentation tests, suggesting that accurate
characterisation of the strain and the time
dependent behaviour of the cheeses was achieved.
The indentation models were also successful in
predicting the steady-state wire cutting force throughthe use of a critical fracture strain criterion.
REFERENCES
ABAQUSs user manual ver 5.8. Hibbitt, Karlssonand Sorensen (UK), Cheshire (1998)
Charalambides MN, Goh SM, Lim SL, JG Williams:The analysis of the frictional effect on stress-straindata from uniaxial compression of cheese, J.Mater. Sci. 36, 2313-2321 (2001)
Goh SM: An engineering approach to food texture
studies. Ph.D. thesis, Imperial College London(2002)
Goh SM, Charalambides MN, Williams JG: Largestrain time dependent behaviour of cheese, J.Rheol., submitted (2002)
Kamyab I, Chakrabarti S, Williams JG: Cuttingcheese with wire, J. Mater. Sci. 33, 2763-2770(1998)
Miller K: Constitutive model of brain tissue suitablefor finite element analysis of surgical procedures,J. Biomech. 32, 531-537 (1999)
Prentice JH, Langley KR, Marshall RJ: CheeseRheology, In Cheese: Chemistry, Physics andMicrobiology, Volume 1, 2
nded. PF Fox, (Ed.)
Chapman and Hall, London, 303-341 (1993)
ACKNOWLEDGEMENTS
The authors would like to thank the BBSRC for
financial support and General Mills for providing the
process cheese.