Professional Documents
Culture Documents
Lava 2012
Lava 2012
Keywords Abstract
Digital Image Correlation, Education, FEA,
Stress Concentrations In this study, we introduce graduate students to an experiment with digital
image correlation (DIC). From an educational point of view, the aim is to
Correspondence
get students familiar with basic aspects of stereovision and DIC, for example,
P. Lava,
speckle pattern, subset size, focus, aperture, etc. First, a homogeneous uniaxial
Department of Mechanical Engineering,
Catholic University College Ghent, tensile test is conducted on a rubber dogbone, allowing the determination of
Association K.U. Leuven, the hyperelastic stress–strain relationship. Next, a nondestructive 2D defor-
Gebroeders Desmetstraat 1, mation test on a rubber specimen with a specific geometry is performed on a
B-9000 Gent, Belgium small tensile setup. At various load steps, images are captured with a low-cost
Email: pascal.lava@kahosl.be webcam and are processed with our in-house DIC software which is available
for each individual student. Assuming a Mooney–Rivlin material model based
Received: June 23, 2011; accepted:
on the determined stress–strain data, one can derive stress concentration fac-
January 10, 2012
tors for the specific holes and notches present in the specimen. A comparison
doi:10.1111/j.1747-1567.2012.00813.x is made to analytical calculations of the stress concentration factors and even-
tually a finite element analysis of the experiment can be performed. In this
way, a synergy between experiment, simulation, and theoretical calculations is
achieved. Moreover, we have accomplished all stages in the engineering pro-
cess of determination of material properties, simulation, and validation. The
particular experimental setup is chosen to avoid the use of special equipment,
for example, sophisticated tensile devices and expensive high-tech cameras,
without losing the focus of the objective.
The main research efforts of the research group The prelab course material consists of a general
Mechanics of Materials, Products, and Processes at this introduction to DIC via a video demonstration. In the
University have been concentrated around the accu- postlab phase students determine stress concentration
racy of this optical–numerical measuring technique2,3 factors via the obtained experimental data, compare
and its application to determine the elastoplastic those to analytical expressions corresponding to the
material parameters by combining experimental data specific geometry and eventually to a finite element
and numerical simulations in a so-called inverse simulation of the experiment.
methodology.4,5 In this contribution, we describe The outline of this paper is as follows. In the next
how we bridge the gap between research and edu- section we briefly describe the prelab phase and the
cation by introducing students to the principles of background of digital image correlation. In section
DIC. A typical lab session consists of three stages, Experimental Setup we focus on the experimental setup
combining a homework assignment (prelab course with specific attention to the material’s choice and the
material review and postlab experimental data analy- test specimen geometry. Next, in section Analysis and
sis, report) and follow-on laboratory experiment. The Report we focus on the data analysis and the report,
introduction of DIC in this scheme is not straightfor- with conclusions in the final section.
ward as we are not only limited in time—a typical
‘‘hands on’’ lab session takes only three hours—but
also we need to accommodate groups of 20 students. Prelab Phase: Introduction to DIC
As such, high-tech experiments combining sophis- As stated above, within the prelab phase students get
ticated tensile machines, expensive cameras, lenses, introduced to the basic principles and mathematics
and commercial software are not an option. The soft- of DIC via a course and a video-demonstration. It is
ware issue was easily solved by using an in-house shown to students that DIC is an optical–numerical
DIC platform MatchID6 that can be provided to each measuring technique offering unique opportunities
student. The next focus was to develop a material test for exploring complex full-field displacement and
that not only produces deformations within the mea- deformation at the surface of objects under any
surement range of DIC (strain ≥ 0.1%), but also does kind of loading in 2D and 3D space. The basic prin-
this at low loading conditions (maximum of 200 N) ciples of the technique are demonstrated in Fig. 1.
to avoid the use of tensile test machines. To minimize Cameras take synchronized images of a specimen in
the cost, the expensive camera and lens equipment the undeformed and deformed state, a stage which
was replaced by a simple webcam. Obviously, this is referred to as the optical part of DIC. Next, these
replacement needed to be validated with respect to images are computationally compared in correlation
accuracy and precision for the developed experiment. software, representing the numerical part. Finally,
Figure 1 Basic principles of digital image correlation: (1) image acquisition of a speckled specimen subject to loading conditions, (2) numerical
comparison of the images via correlation software, and (3) full-field deformation fields.
Figure 3 Geometry of the specimen. Dimensions are in mm. Various holes and notches are involved, including a circular hole (centre), an u-notch
(right-centre), a semicircular notch (left-centre), an elliptical hole with major axis parallel to the loading direction (upper-right), an elliptical hole with
major axis perpendicular to the loading direction (lower-right), a rectangular hole and rows of circular holes parallel and perpendicular to the axial
loading direction.
factors. Accordingly, the tensile test is performed speckle patterns involved—subset dimensions within
on a rubber specimen with the specific geometry the range of 19 × 19 to 35 × 35 pixels2 will result in
shown in Fig. 3. The various holes are chosen in a successful correlation.
order to allow different student groups to deal with After accomplishing a good experimental setup,
different stress concentrations. In addition, for these students can embark on the loading of the specimen.
specific holes analytical formulas can be used in order First, a reference image of the unloaded specimen
to validate the experimental results10,11 with the is taken. Next, the specimen is loaded with a force
assumption that the holes do not influence each other. of roughly 20 N and a first deformed image is cap-
Each experimental setup is equipped with a tured. A preliminary correlation run is performed
Logitech WebCam Pro9000 which has a resolution of and the analysis of a correlation coefficient pattern
1600 × 1200 pixels. Within a regular DIC experiment as depicted in Fig. 4 gives the students a good idea
the setup of the camera is of the utmost importance of the performance of their test setup, problems with
and should be handled carefully by the students. the speckle pattern, overexposure, shadowing, etc.
Not only good focus should be guaranteed, but The exact magnitude of the weight can be deter-
also the distance to the specimen must be set so that mined via the measurement on a tension bar using
the specimen roughly fills the entire field of view. strain gauges in a full-bridge circuit, representing a
If the specimen is larger than the field of view, we manufactured load cell (see Fig. 2). As such, students
lose data at the edges, whereas a too small specimen also get familiar with the concept of a load cell com-
will hamper our spatial resolution. In addition, the pensating for various influences on the strain gauge
object must remain in the field of view for the entire configuration, for example, temperature raise by the
test. Accordingly, the field of view should be adjusted lamp and bending effects on the tensile bar induced
properly. A suitable setup results in pixel dimensions by the experimental setup. This allows a very accurate
of ≈0.125 mm. This means that—for the various determination of the load step under consideration.
Figure 4 The DIC correlation coefficient r ( r ∈ [0, 1] with 1 corresponding to a 100% match) corresponding to F = 100 N. For clarity, the applied load
is shown.
The abovementioned steps are repeated for additional potential of first order. This model is based on strain
weights of approximately 20 N up to a maximum invariants which are functions of the three principal
loading condition of 200 N. Accordingly, the students stretch ratios λ:
have a reference image and 10 deformed images with
corresponding loading conditions which they can use I1 = λ21 + λ22 + λ23 (1)
in the analysis phase. I2 = λ21 λ22 + λ22 λ23 + λ21 λ23 (2)
I3 = λ21 λ22 λ23 . (3)
Analysis and Report
The Mooney–Rivlin strain energy potential of first
In the analysis phase, the primary goal is to determine
order is then given by:
the stress concentration factors around specific holes
at the different load steps. To do this, students need U = C10 (I1 − 3) + C01 (I2 − 3) + (I3 − 1)2 /2D, (4)
to determine the displacements as close as possible to
the edge of the hole. This is done by selecting a subset with C10 , C01 and D are constants that can be obtained
size as small as possible (lower-limited by the gran- from curve fitting the stress–strain relationship of the
ularity of the speckle-pattern), and a data resolution homogeneous uniaxial tensile test conducted on a
as high as possible (small step size). Next, the corre- rubber dogbone at the start of the hands-on phase
lation software allows the determination of the strain (see section Experimental Setup). If available, a stan-
components as shown in Fig. 5. As can be inferred, dard finite-element simulation software package12
the requirement on the magnitude of the determined converts the stress–strain data into C10 , C01 , and D.
strain (≥0.1%) is clearly fulfilled. The aim is, how- If not, students can adopt the following ‘‘default”
ever, to determine stresses. As a result, the students values which result from a previously performed
need to provide the software with a material model stress–strain determination:
in order to convert strains into the corresponding
stresses. C10 = −7.62 E05 J,
Hyperelastic material models are based on strain C01 = 2.19 E06 J,
energy density functions, usually denoted by U. We
adopt a Mooney–Rivlin model with a strain energy D = 3.56 E − 08 J−1 .
Accordingly, the stress components can be obtained loaded in tension.10 This concentration factor depends
as: on the radius of the hole r = 8 mm and the width
of the plate D = 170 mm. These values are obtained
2 ∂U ∂U ∂U
S = DEV + I1 B− B·B , (5) by subtracting the sizes of the left and right u-notch.
J ∂I1 ∂I2 ∂I2
Substituting these values into the formula of Ref. 10:
with B the left Cauchy–Green strain tensor. 2 3
Students can directly convert the DIC determined 2r 2r 2r
Kt = 3.00 − 3.13 + 3.66 − 1.53 ,
strains into stresses via our in-house software and D D D
the obtained values of C10 , C01 , and D. This yields a (7)
longitudinal stress component as displayed in Fig. 6.
Accordingly, students are able to embark on the study yields Kt = 2.736. In Fig. 7 the evolution of the stress
of stress concentration factors for the specific holes concentration factor measured with DIC at a maxi-
involved. Below we sketch the procedure for the mum load of 200 N is displayed. As can be inferred,
central hole, identical conclusions can be drawn for for this particular load step and very close to the
the other ones. central hole the stress concentration factor is in good
The (net) static stress concentration factor in agreement with the theoretical prediction, whereas
the elastic range, Kt , is defined as the ratio of the the stress concentration factor drops quickly down to
maximum stress, σmax , to the (net) nominal stress, 1 when one moves away from the centre of the hole.
σnom 11 : If students have access to finite-element simulation
software they can perform a finite-element analysis
σmax
Kt = . (6) (FEA) of the considered experiment and validate this
σnom numerical model. The hyperelastic material model is
As such, with the specific force and the geometry identical to the one used to convert DIC strain fields
of the specimen, students can determine the nomi- into stresses, that is, a Mooney–Rivlin model based
nal stress. Using the DIC determined maximum stress on the stress–strain relationship of the homogeneous
close to the central hole, the DIC measured stress con- uniaxial tensile test conducted on a rubber dogbone.
centration factor KtDIC is found. Alternatively, one can Next, the load steps are determined in accordance
analytically determine the static stress concentration to the values measured with the tension bar. In
factor for a plate containing a centrally located hole Fig. 8 the results of a finite element simulation for
Figure 6 DIC results of the longitudinal stress component σL corresponding to F = 100 N. Units are in Pa.
2.5
the analytical values. As such, the bulk of the devia-
2
tions between FEA and analytical calculations can be
1.5 ascribed to the nonlinear material behaviour of rubber
0 25 50 75 100 125 150 175 200
and to the assumption that the holes do not influence
F (N)
each other. This makes that the FEA results in Table 1
(full geometry and nonlinear material behaviour)
Figure 9 The maximum longitudinal stress component σL (upper panel)
should be considered as the benchmark ones. The
and the derived stress concentration factor (lower panel) for various
load steps. The circles and the triangles represent the results of the FE DIC results have a standard deviation that is ≈13%
simulation and the DIC measurement, respectively. The solid line is the of the mean value, which is fair in view of the basic
theoretical prediction. setup involved.
Accordingly, we can conclude that the chosen
approximations are justified for a student test. A
the complexity of the measurement errors involved.
more accurate measurement would definitely involve
The mutual comparison, however, lies always within
a more fine-tuned experimental setup and a higher
the 10% range, which is fair for a basic student’s
resolution camera. Moreover, one would also need
laboratory test.
to take into account that the theoretical predictions
Table 1 summarizes the analytical values of the
are only valid in the linear elastic range, whereas we
stress concentration factor for the specific holes and
are dealing with a nonlinear hyperelastic material.
notches. In addition, the mean values and the stan-
The primary goal of the lab course, however, is to
dard deviations of the DIC and FEA results are listed.
introduce students with DIC measurements and the
In general, the FEA and DIC results correspond well,
various aspects involved with this technique.
except for the holes with the smallest radii. This could
In the final report, students create and present the
be because of the fact that it is very hard to cut this
diagrams displayed in Figs. 7–9 for a particular hole of
particular notch out of the rubber specimen, inducing
the specimen (i.e. circular hole, u-notch, semicircular
locally larger stress concentrations. The comparison
u-notch, elliptical hole, row of holes parallel to axial
to the analytical values is reasonable. A FEA con-
tension, row of holes normal to axial tension, or a
vergence study is performed, excluding the impact
rectangular hole with rounded corners). They should
of the element mesh size. Next, it should be noted
critically approach the obtained results and draw con-
that the analytical expressions in Refs. 10 and 11 are
clusions with respect to the underlying assumptions
and the experimental setup. In the end, it is the pur-
Table 1 Summarizing table containing the analytical stress concentra- pose that the student has a basic understanding of DIC
tion factor (second column) for each type of geometry present in the and how to use it in an experimental environment.
specimen
FE DIC
Geometry Ktan Kt σ KtFE Kt σ KtDIC
Conclusions
Circular hole 2.74 2.49 0.045 2.52 0.225
A recently developed lab course on digital image
U-notch 2.26 2.54 0.056 2.59 0.237
Semi-circular notch 2.3 2.03 0.003 1.97 0.207
correlation measurements in the classroom was out-
Elliptical (a F ) 1.87 1.81 0.003 1.72 0.303 lined. The lab course can accommodate a maximum
Elliptical (a ⊥ F ) 2.97 2.66 0.055 3.11 0.306 of 20 students in groups of two. To accomplish this,
Rectangular 2.5 1.87 0.021 1.88 0.342 an experimental setup is chosen that avoids the use
Row of circles ( F) 2.07 2.00 0.006 1.65 0.248 of special equipment like sophisticated tensile test
Row of circles (⊥ F) 2.5 2.24 0.002 2.06 0.182
machines and expensive high-tech cameras. Uniaxial
Column number three and four (five and six) represent the FEA (DIC) tensile tests on a rubber sheet specimen are performed
mean values and standard deviations over the different load stepsx. For on a small experimental setup that can be loaded
the elliptical geometries, a corresponds to the major axis of the ellipse. manually. During the test, images are captured with
a low-budget webcam and are processed with our Material Behaviour through Inverse Modelling and
in-house DIC software available for each student. As DIC,’’ Experimental Mechanics 48:421–433 (2008).
a final result, experimentally determined stress con- 5. Coppieters, S., Cooreman, S., Sol, H., Van Houtte, P.,
centration factors around the holes in the specimen and Debruyne, D., ‘‘Identification of the
can be compared to the results of a finite element Post-Necking Hardening Behaviour of Sheet Metal
simulation and the analytical predictions. This cre- by Comparison of the Internal and External Work in
ates a lab exercise which combines an experiment, an the Necking Zone,’’ Journal of Materials Processing
analytical solution and a finite-element validation. Technology 211:545–552 (2011).
6. MatchID, Web site: http://www.matchid.org
[accessed on 25 January 2012].
References 7. Sutton, M., Wolters, W., Peters, W., Ranson, W.,
and McNeill, S., ‘‘Determination of Displacements
1. Avril, S., Bonnet, M., Brettele, A-S., Grediac, M.,
Using an Improved Digital Correlation Method,’’
and Hild, F., ‘‘Overview of Identification Methods of
Image and Vision Computing 1:133–139 (1983).
Mechanical Parameters Based on Full-field
8. Chu, T., Ranson, W., and Sutton, M., ‘‘Applications
Measurements,’’ Experimental Mechanics 48:381–402
of Digital-Image-Correlation Techniques to
(2008).
Experimental Mechanics,’’ Experimental Mechanics
2. Lava, P., Cooreman, S., Coppieters, S., De Strycker,
25:232–244 (1985).
M., and Debruyne, D., ‘‘Assessment of Measuring
9. Sutton, M., Orteu, J., and Schreier, H., Image
Errors in DIC Using Deformation Fields Generated
Correlation for Shape, Motion and Deformation
by Plastic FEA,’’ Optics and Lasers in Engineering
Measurements, Springer, USA (2009).
47:747–753 (2009).
10. Young, W., and Budynas, R., Roark Formulas for Stress
3. Lava, P., Cooreman, S., and Debruyne, D., ‘‘Study of
and Strain, McGraw-Hill, USA (2002).
Systematic Errors in Strain Fields Obtained Via DIC
11. Pilkey, W., Peterson’s Stress Concentration Factors, John
Using Heterogeneous Deformation Generated by
Wiley & Sons, USA (1997).
Plastic FEA,’’ Optics and Lasers in Engineering
12. Abaqus, Theory Manual, Version 6.7 (2007).
48:457–468 (2010).
4. Cooreman, S., Lecompte, D., Sol, H., Vantomme, J.,
and Debruyne, D., ‘‘Identification of Mechanical