Optics and Lasers in Engineering 100 (2018) 248–258

Contents lists available at ScienceDirect

Optics and Lasers in Engineering

journal homepage: www.elsevier.com/locate/optlaseng

Photoelastic method to quantitatively visualise the evolution of whole-field

stress in 3D printed models subject to continuous loading processes
Yang Ju a,b,∗, Zhangyu Ren c, Li Wang c, Lingtao Mao a, Fu-Pen Chiang d
a
State Key Laboratory of Coal Resources and Safe Mining, China University of Mining and Technology at Beijing, D11 Xueyuan RD, Beijing, 100083, China
b
State Key Laboratory for Geomechanics and Deep Underground Engineering, China University of Mining and Technology at Xuzhou, 1 University Ave, Xuzhou,
221006, China
c
School of Mechanics and Civil Engineering, China University of Mining & Technology at Beijing, D11 Xueyuan RD, Beijing, 100083, China
d
Department of Mechanical Engineering, Stony Brook University, NY, 11794-2300, USA

a r t i c l e i n f o a b s t r a c t

Keywords: The combination of three-dimensional (3D) printing techniques and photoelastic testing is a promising way to
Continuous whole-field stress quantitatively determine the continuous whole-field stress distributions in solids that are characterized by com-
Visualization plex structures. However, photoelastic testing produces wrapped isoclinic and isochromatic phase maps, and
Photoelastic testing
unwrapping these maps has always been a significant challenge. To realize the visualization and transparenti-
Fringe orders
zation of the stress fields in complex structures, we report a new approach to quantify the continuous evolution
3D printing
Frame processing of the whole-field stress in photosensitive material that is applicable to the fabrication of complex structures
using 3D printing technology. The stress fringe orders are determined by analyzing a series of continuous frames
extracted from a video recording of the fringe changes over the entire loading process. The integer portion of the
fringe orders at a specific point on the model can be determined by counting the valleys of the light intensity
change curve over the whole loading process, and the fractional portion can be calculated based on the cosine
function between the light intensity and retardation. This method allows the fringe orders to be determined from
the video itself, which significantly improves characterization accuracy and simplifies the experimental operation
over the entire processes. To validate the proposed method, we compare the results of the theoretical calcula-
tions to those of experiments based on the diametric compression of a circular disc prepared by a 3D printer with
photosensitive resin. The results indicate that the method can accurately determine the stress fringe order, except
for points where the deformation is too large to differentiate the fringes pertaining to photoplasticity.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction of the dynamic evolutionary processes related to the whole field stress
is the key to understanding the mesoscopic damage, deformation, and
The ability to accurately characterize and visualize the dynamic failure mechanisms in underground engineering problems.
stress distribution in solids that are characterized by complex structures To accomplish these goals, a variety of techniques are used that
is of vital significance in solving a variety of engineering and technical can be roughly classified as in-situ monitoring [10,11], laboratory tests
problems. For example, in natural resource exploration for oil, natural [12–14], and numerical simulations [15–17]. In terms of experimentally
gas, and coalbed methane, the continuous evolution of the stress field quantifying the stress field that essentially governs the deformation and
of the underground rock must be mapped to properly evaluate the effi- failure of a complex system, photoelasticity is an optical method that
ciency of reservoir stimulation and the production rate [1–3]. The ability uses fringe contours to indicate the principal stress difference and direc-
to resolve problems in geological and environmental engineering, such tion, and is widely used for two-dimensional (2D) and three-dimensional
as geological waste disposal, geological storage of CO2 , and geothermal (3D) stress analyses in mechanical engineering [18,19], medical science
energy utilization, is highly dependent on knowledge of the dynamic [20,21], rock mechanics and rock engineering [22–24], etc. However,
evolution of the underground stresses [4–7]. It is crucial to understand even though it is widely applied, this method usually suffers from two
the continuous stress distribution of each component in mechanical en- extremely difficult problems: difficulties in preparing the required com-
gineering applications to properly evaluate the mechanical properties plex models, and challenges in quantitatively characterizing the whole-
of complex systems [8,9]. For these applications, direct quantification field stress. However, the recent advent of 3D printing technology has
simplified the process of preparing complex models, so this is no longer

∗
Corresponding author.
E-mail addresses: [email protected], [email protected] (Y. Ju).

http://dx.doi.org/10.1016/j.optlaseng.2017.09.004
Received 20 May 2017; Received in revised form 29 July 2017; Accepted 6 September 2017
Available online 14 September 2017
0143-8166/© 2017 Elsevier Ltd. All rights reserved.
Y. Ju et al. Optics and Lasers in Engineering 100 (2018) 248–258

an insurmountable difficulty. Studies show that it is possible to choose Table 1

Basic mechanical properties of 3D printing materials.
photosensitive resin as the printing material, which exhibits birefrin-
gence phenomena, a basic precondition of performing photoelastic test- Materials ƒc (MPa) ƒt (MPa) E (GPa) 𝜈
ing during loading [22–24]. 25 °C Vero Clear 81.6 38.1 3.10 0.38
To date, numerous methods have been developed to determine the
Note: fc and ft refer to the uniaxial compressive strength and direct ten-
whole-field stress. One approach is the phase shift technique, which is
sile strength, E refers to the elasticity modulus, and 𝜈 refers to Poisson’s
widely used in data acquisition applications. This approach includes a ratio.
variety of methods [25–29], such as the ten-step phase shift method
[30] or the improved six-step phase shift method [31]. In the latter,
the four- and six-step methods are combined to determine the principal
stress direction (isoclinic) and principal stress difference (isochromatic).
This process requires ten images to be obtained while adjusting the plane
and circular polariscope, although the accuracy of the method is very
reliant on the quality of the unwrapping. Another approach is the three-
fringe photoelasticity or RGB photoelasticity [32–34] method, which
requires only one color image to estimate the whole field fringe order,
and is based on a calibration table of color components to match the red,
green, and blue components at various points on the model. In this way,
the whole field principal stress difference can be determined; however,
the calculation accuracy can be influenced by a variety of factors, such
as the spectral composition and stability of the light source, the trans-
mission response of the polariscope components, the error of the quarter
wave plates, etc. To overcome these shortcomings, several approaches
have been proposed to upgrade the RGB method using advanced ex-
perimental equipment to handle cases where the fringe order is higher
Fig. 1. Photograph of the experimental apparatus used to visualize and capturing the
than three [35–40]. Still another approach is the step-loading method
fringe patterns.
[41,42], which is based on the theory of phase shifting. This method
keeps a record of the light intensities of different optical systems at a
certain load level, and then records the data again when the load in- physical/mechanical parameters of the printing material are shown in
creases. One advantage of this method is that it can be used to compute Table 1. The arrangement of dark-field circular polariscopes used in the
the fringe orders based on the load history, but a limitation is that it experiments is shown in Fig. 1. A digital servo-control universal test-
cannot be used over the duration of a continuous loading process. Fi- ing machine with a loading capacity of 100 kN and a loading rate of
nally, the Fourier transform technique can be used to estimate the whole 0.5 mm/min was used to exert continuous compressive pressure on the
field fringe orders, but many images are required [43]. Even though disc, a digital camera was employed to record the fringe contours dur-
some areas of research, such as determining the mining induced stress ing the loading process, and a monochromatic light filter was installed
in surrounding rock during tunnel excavations, require an understand- in the lens of the digital camera for clear observation.
ing of the dynamic evolutionary process of the whole field stress during
continuous loading [44–46], as explained above, existing photoelastic- 2.2. Isochromatic quantification determination method
ity methods cannot determine the whole field stress during continuous
loading processes, and the accuracy of the calculated results is greatly The experimental setup, including the polarizer, two quarter wave
limited by the complexity of the unwrapping methods. Therefore, it is plates, and analyzer, is shown in Fig. 2. The angles of the optical axes
desirable to develop a new method to quantify the whole field stress of the polarizer and analyzer, and the fast axes of the two quarter wave
during continuous loading processes. plates with respect to the vertical direction are also shown in the fig-
In this paper, we propose a new method for estimating the entire ure. The relationship between the light intensity and retardation can
principal stress difference for the elastic stage in a continuous loading be derived according to the matrix theory of photoelasticity [47]. The
process. The method utilizes a 3D printer to prepare experimental mod- normalized Jones vector of the polarized light from the polarizer can be
els with transparent photosensitive resin. Then, a circular transparent expressed as:
disc in diametric compression is used to visualize and record the whole- ( )
0
field fringe contours using a camera. Finally, the integer portion of the 𝑃𝑝 = (1)
1
fringe orders at a specific point on the model are determined by count-
The Jones matrix of the quarter wave plates, where the angle be-
ing the valleys of its light intensity change curve over the whole load-
tween their fast axes and the Y-axis are ±45°, can be written as:
ing process, and the fractional portion is calculated based on the cosine ( )
function between the light intensity and retardation. To validate the 𝑖+1 1 𝑖
𝑄45 = . (2)
proposed method, we discuss the primary sources of error, and compare 2 𝑖 1
the results of the theoretical calculations and those of the experiments ( )
𝑖+1 1 −𝑖
based on the diametric compression of a circular disc. The results in- 𝑄−45 = . (3)
dicate that the proposed method can accurately determine the whole 2 −𝑖 1
field fringe order, except for points where the loading deformation is For the specimen, the retardations of the fast and slow axes are rep-
too large to distinguish the fringes pertaining to the photoelasticity. resented by 𝛿, the angle of the fast axis to the X-axis is represented by
𝛽, and its Jones matrix can be given as:
2. Materials and methods ( 𝑖𝛿 2 )
𝑒 cos 𝛽 + sin2 𝛽 (𝑒𝑖𝛿 − 1) sin 𝛽 cos 𝛽
𝑅= 𝑖𝛿 𝑖𝛿 . (4)
(𝑒 − 1) sin 𝛽 cos 𝛽 𝑒 sin 𝛽 + cos 𝛽
2 2
2.1. Model preparation and testing
The Jones matrix of the analyzer can be written as:
( )
The circular transparent disc model was prepared using an Object 1 0
Connex 500 3D printer and the Objet Studio software package. The 𝑃𝑎 = . (5)
0 0

249
Y. Ju et al. Optics and Lasers in Engineering 100 (2018) 248–258

Fig. 2. Diagram showing the arrangement of the circular polariscope, analyzer, and quarter wave plates.

In the Jones matrix method, the Jones vector 𝛼 of the light coming the relationship between the light intensity and the retardation is a co-
from the analyzer can be calculated by: sine function, hence the light intensity of each point in every frame can
be regarded as the value of this cosine function. Furthermore, the light
𝛼 = 𝑃𝑎 𝑄−45 𝑅𝑄45 𝑃𝑝 (6)
intensity change curve directly indicates the retardation at each point. In
Considering Eqs. (1–5), the Jones vector 𝛼 can be given by: the dark field of the circularly polarized light, the darkest fringe repre-
( ) ( )( 𝑖𝛿 2 ) sents the integer portion of the fringe orders, the lightest fringe denotes
1 0 𝑖+1 𝑖 −𝑖 𝑒 cos 𝛽 + sin2 𝛽 (𝑒𝑖𝛿 − 1)sin𝛽cos𝛽
𝜶 = 𝑖𝛿 𝑖𝛿
one half of the fringe orders, and the valleys and peaks in the curve of
0 0 2 −𝑖 1 (𝑒 − 1)sin𝛽cos𝛽 𝑒 sin 𝛽 + cos 𝛽
2 2
Fig. 4(b) represent the dark and light fringes of Point 1 in the loading
( )( )
𝑖+1 1 𝑖 0 process. In addition, the number of valleys denotes the number of dark
. (7)
2 𝑖 1 1 fringes between the beginning of the load to the stage we want to cal-
culate, and is the integer part of the fringe orders. As for the fractional
After completing the calculation, 𝛼 can be written as: part of the fringe orders, this can be calculated from the cosine function
( )
1 𝑒𝑖𝛿 − 1 based on the values of the last valley and peak in the curve.
𝛼= . (8)
2 0
The intensity of the light expressed by Eq. (8) is: 2.3.1. Determination of the integer portion of the fringe orders
As mentioned previously, the number of valleys indicates the inte-
𝛿
𝐼 = 𝛼𝛼 = sin2 . (9) ger part of the fringe orders for the point that will be calculated. Con-
2
sequently, the number of valleys should be carefully determined. The
In the test, if the light intensity of source Is and the background in-
light intensity changes of Point 1 in Fig. 4(a) are used as an example to
tensity Ib are considered, the light intensity in Eq. (9) can be written as:
explain the method of determining the number of valleys. As shown in
1 Fig. 5, the wave signal of Point 1 is translated into a filter curve based
𝐼 ′ = 𝐼𝑏 + 𝐼 (1 − cos𝛿). (10)
2 𝑠 on a value Im near the average of the first peak Ip and valley Iv . That is,
This equation indicates that the relationship between the light in- if the light intensity is larger than the value Im , then the light intensity
tensity I’ and the retardation 𝛿 can be described by a cosine function. is equal to the first peak value Ip . Otherwise, the light intensity equals
In other words, the light intensity at a certain point on the model will the first valley Iv . To ensure they are different, we can use any value
be the cosine of 𝛿 with the load increment for the duration of the load- to replace the first peak and valley, so the result of the translation is
ing process. On this basis, we propose a method to determine the fringe shown as a filter curve (shown as the red curve in Fig. 5). It is worth
orders of a point at all stages of the loading process. noting that the rate of light intensity near Im is larger than the points
that are close to the peak and valley in that curve, so the error caused
2.3. Stress field quantification by the number of points Pi will decrease, i.e., the crossover point of the
line of light intensity that equals Im and the curve of the light intensity
The process for determining the fringe orders is as follows. First, the changes. Fig. 6 shows a flowchart of the method used to determine the
changes in the fringe orders during the loading process are captured integer part of the fringe orders. First, the light intensity curve of each
by video camera. Then, the video is segmented into multiple intervals point is translated into a filter curve. Then, based on the determination
(frames) of equal length, and the color frames (filtered by green fil- that if the light intensity of any point in the filter curve I(i) satisfies:
ter) are translated into gray images by averaging the values of the red,
𝐼 (𝑖 ) = 𝐼 𝑣 , (11)
green, and blue pixels, as shown in Fig. 3. The light intensity changes
of one pixel over the course of the entire loading process forms a one- and the light intensity of next point in filter curve I(i + 1) satisfies:
dimensional (1D) sequence. Therefore, the light intensity changes of all
pixels in all frames forms a 3D matrix that includes all the informa- 𝐼 (𝑖 + 1) = 𝐼𝑝 , (12)
tion about the changes in the fringe orders. As shown in Fig. 4(a), as then the number n of points that satisfy the condition can be counted
the frame number increments, the 2D image matrixes form a 3D matrix using the process described in the flowchart, as illustrated by points P1
where the values for each point in the frame number incremental direc- to P9 in Fig. 4, because the relationship between the number n and the
tion indicate the light intensity changes over the whole loading process, number of valleys in the light intensity curve nv is:
as shown in Fig. 4(b). The curve in the figure presents the light intensity
changes of Point 1 on the disc as the load increments. Based on Eq. (10), 𝑛𝑣 = 𝑛 − 1. (13)

250
Y. Ju et al. Optics and Lasers in Engineering 100 (2018) 248–258

(a) (b)
Fig. 3. Translation from the color figure into a gray image: (a) monochromatic, and (b) grayscale isochromatic fringe pattern of the circular disc under a diametric compression of 2 kN.

Fig. 4. Light intensity changes during the loading process: (a) the frames used to form a
three-dimensional matrix, and (b) the light intensity changes of Point 1 during the loading
process.

Fig. 6. Flowchart showing the procedure for determining the integer part of the fringe
orders.

Therefore, the number of valleys in the light intensity curve is the

same as the integer part of the fringe orders.

2.3.2. Determination of the fractional portion of the fringe orders

Based on Eq. (10), the relationship between the light intensity of each
point and the retardation of the same point during the loading process
can be described by a cosine function. The retardation can therefore be
written as:
( )
2(𝐼 ′ − 𝐼𝑏 )
Fig. 5. Description of the method used to determine the number of valleys. 𝛿 = arccos 1 − , 𝛿 ∈ [0, 𝜋]. (14)
𝐼𝑠

251
Y. Ju et al. Optics and Lasers in Engineering 100 (2018) 248–258

Fig. 7. Light intensity of the three points shown in Fig. 4(a) with the increment of the
frame number.

Fig. 8. Light intensities of the peaks and valleys for the three points shown in Fig. 7.

If the background light and light source intensity do not change, the
cosine function can also be used to describe the peak and valley values:

𝐼𝑝 + 𝐼𝑣 𝐼𝑝 − 𝐼𝑣
𝐼′ = − 𝑐𝑜𝑠𝛿, (15)
2 2
then the arccosine function can be written as:
( )
δ = arccos (𝐼 ′ − 𝐼𝑎𝑣 ∕(𝐼𝑝 − 𝐼𝑎𝑣 )),
or Fig. 9. Illustrations of three situations used to calculate the fractional part of the fringe
( ) orders: (a) the fringe orders of the rest are less than half, (b) the fringe orders of the rest
δ = arccos (𝐼 ′ − 𝐼𝑎𝑣 ∕(𝐼𝑎𝑣 − 𝐼𝑣 )), δ ∈ [0, π], (16)
are less than one and a half, but more than one, and (c) the fringe orders of the rest are
less than one but more than half.
𝐼𝑝 + 𝐼𝑣
𝐼𝑎𝑣 = , (17)
2
It is noteworthy that both of the equations that make up Eq. (16) are
used to calculate the retardation of the three cases shown in Fig. 9. The the values of the last peak and valley need to be determined for each
first was used for the case of Fig. 9(c), while the second was applied to point, and the number of peaks and valleys should be considered.
the cases of Figs. 9(a) and (b). Owing to instabilities and inequalities in According to the method used to determine the integer part of the
the light sources, manufacturing errors in the lenses used to create cir- fringe orders, the fringe orders before the last valley can be determined.
cularly polarized light, light refraction caused by the surface roughness However, for the part after the last valley, there are three cases that must
of the specimen etc., the values of the peaks and valleys for the same be evaluated to calculate the corresponding phase angle 𝛿. As shown in
fringe order may be different for points in different positions. For exam- Fig. 9(a–c), point Pl in these figures is the last point that can be counted
ple, when the peak and valley values of the three curves in Fig. 7 that by the method used to calculated the integer part of the fringe orders,
represent the light intensity change curves of the three points in Fig. 4(a) so the fringe orders before the valley closet the point Pl can be deter-
are counted, we can see that the peak and valley values are different for mined and the rest (see the gray parts in Fig. 9(a–c)) are regarded as
the same number of peaks or valleys. Consequently, the peak and valley the fractional parts, which can be calculated using the values of the last
values used for the calculation of retardation for the same fringe order peak and valley. For the three figures in Fig. 9, the last peak and valley
in different points are also different. At the same time, the peak and marked in these figures are close to the last point in these curves; thus,
valley value of the same point are also different for different numbers they are the best choice to use with the cosine function to calculate the
of peaks and valleys. As shown in Fig. 8, the peak values decrease as retardation of the last point. As shown in Fig. 9, I’p is the light intensity
the number of peaks increases, but the valley values increase. To deter- of the last peak, and I’v is the light intensity of the last valley. Then,
mine the fractional part of the fringe orders using the cosine function, I’p and I’v can be substituted into Eq. (16), and the retardation in the

252
Y. Ju et al. Optics and Lasers in Engineering 100 (2018) 248–258

Fig. 12. Sketch describing the change trend of the first peak and valley values.
Fig. 10. Three cases before the second point based on the flowchart shown in Fig. 6.
Table 3
Table 2 Calculation of the regions where the light intensity curve changes before the
Calculation formulas of the fractional part nd in the three simulations. second point in Fig. 5.

Judging condition Δ Fractional part nd Item Ip Iv 𝛿 N

δ δ
(a) 𝐼 ′𝑙 ≥ 𝐼 ′𝑝 &𝐼 ′ ′𝑙 ≥0 arccos(−(𝐼 − 𝐼𝑎𝑣 )∕(𝐼𝑎𝑣 − 𝐼𝑣 ))
′
Case 2 Ip 𝐼𝑣′ + (𝐼𝑝 − 𝐼𝑝′ ) arccos((𝐼 − 𝐼𝑎𝑣 )∕(𝐼𝑝 − 𝐼𝑎𝑣 )) + π
′
𝑙 2π 2π
δ δ
(b) 𝐼 ′ ′𝑙 ≤ 0 arccos(−(𝐼 ′ − 𝐼𝑎𝑣 )∕(𝐼𝑎𝑣 − 𝐼𝑣 )) +1 Case 3 Ip Iv arccos(−(𝐼 ′ − 𝐼𝑎𝑣 )∕(𝐼𝑎𝑣 − 𝐼𝑣 )) +1
2π 2π
δ
(c) 𝐼 ′𝑙 ≤ 𝐼 ′𝑝 & 𝐼 ′ ′𝑙 ≥ 0 arccos((𝐼 ′ − 𝐼𝑎𝑣 )∕(𝐼𝑝 − 𝐼𝑎𝑣 )) + 𝜋
𝑙 2π
Note: I’pl is the light intensity of point Pl , I’l is the light intensity of the last point in the
curve, and I’’l is the light intensity change rate of the last point, which is represented by However, there are some regions where the peaks and valleys can-
the change rate of the last point in the curve that is formed by quadratic fitting about not be determined by the method introduced in Fig. 6 (see the region
ten points before the last point. before point P2 in Fig. 5). In these regions, there are no valleys or peaks;
hence, the cosine function cannot be determined. Instead, to calculate
the fringe orders in these regions, we choose the first peak and valley
scale [0, 𝜋] for the three simulations in Fig. 9 can be calculated using value in another region to substitute for the peak and valley of the point
the equations in Table 2. where the cosine function must be determined. However, due to the dif-
As mentioned previously, the integer and fractional parts of the ferences in the values of the peaks or valleys at different points, some
fringe orders have been determined, hence the fringe orders N can be error will exist in the retardation calculation. To improve the accuracy
calculated by: of the calculation, we divide it into three cases, as seen in the sketch in
𝑁 = 𝑛𝑣 + 𝑛𝑑 , (18) Fig. 10. Case 1 represents the region before the first peak, Case 2 repre-
sents the region before the first valley, and Case 3 is the region before
where nv , nd are the integer part and fractional part of the fringe orders, the point that was determined by the flowchart in Fig. 6 and closest to
respectively. the first valley (see P2 in the Fig. 5).

Fig. 11. Depictions of the first gray image and the light intensity changes in two directions: (a) the distribution of the initial light intensity, (b) the light intensity changes in the horizontal
and vertical directions, (c) the first peak and valley value changes with the initial light intensity in the vertical direction, and (d) the first peak and valley value changes with the initial
light intensity in the horizontal direction.

253
Y. Ju et al. Optics and Lasers in Engineering 100 (2018) 248–258

Fig. 13. Photoelastic fringe patterns selected from the loading process: (a) the frame when the loading pressure was 1 kN, (b) the frame when the loading pressure was 2 kN, and (c) the
frame when the loading pressure was 3 kN.

Fig. 14. Comparison of the experimental and theoretically calculated results: (a) experimental results when the pressure was 1 kN, (b) experimental results when the pressure was 2 kN,
(c) experimental results when the pressure was 3 kN, (e) theoretically calculated results when the pressure was 1 kN, (f) theoretically calculated results when the pressure was 2 kN, and
(g) theoretically calculated results when the pressure was 3 kN.

Table 4 trend for the first peak and valley were similar; hence, we assume that
Results of the fringe values in the different stages.
the influence of the initial light intensity was primarily focused on the
1 kN 2 kN 3 kN average background, that is the Ib in Eq. (10), which means the difference be-
Center fringe orders 1.39 2.54 3.87
tween the first peak and valley was constant. As shown in Fig. 12, there
Fringe value 36.66 40.12 39.5 38.76 are three cosine functions with different peaks and valleys. That means
that, if we want to express y2 or y3 with y1 , we only need to add a
constant.
Therefore, the method that used the other peaks and valleys to cal-
culate the retardation for the regions where the peaks and valleys could
The influence of the non-uniform light source can be observed in
not be determined must be adjusted. At present, we are not able to re-
the frames before loading, as shown in the Fig. 11(a). To explain this
vise the method used for Case 1 in Fig. 11; however, the corrections for
more clearly, the light intensity in the horizontal and vertical lines was
Cases 2 and 3 are summarized in Table 3. The conditions for these cor-
counted, as shown in Fig. 11(b), because the changes in the light in-
rections are similar to those for the method to calculate the fractional
tensity were evident. Then, 19 points were selected in the two lines,
part, so they are not shown.
respectively, to show the influence of the initial light intensity to the
It is worth noting that there are some errors and limitations in this
first peak and valley value. The relationship between the initial light in-
approach. In later sections, these will be discussed and a new method
tensity and the first peak and valley values are described by the curves in
for calculating the fringe orders will be presented.
Figs. 11(c) and (d). These curves indicate that the light intensity change

254
Y. Ju et al. Optics and Lasers in Engineering 100 (2018) 248–258

Fig. 16. Changes in the absolute-value of the difference in the fringe orders for two meth-
ods versus the increment of displacement: (a) the calculated points in the range of y > 0,
and (b) the calculated points in the range of y < 0.

Fig. 15. Detailed comparison of the experimental and theoretically calculated results: (a)
comparison of the principal stress difference in the horizontal direction, and (b) compar-
ison of the principal stress difference in the vertical direction.
where 𝜎 1, 𝜎 2 are the principal stresses. In this equation, the theoretical
principal stress difference in the circular disc under diametric compres-
sion can be calculated. Hence, the accuracy of the experimental calcu-
3. Result and discussion lation results can be verified by comparison with the theoretical results
once the photoelastic fringe values have been determined. In the center
3.1. Diametric compression of a circular disc of the circular disc (𝑥 = 0, 𝑦 = 0), the principal stress difference is:
4𝑃 𝑁𝑓
Based on the theoretical resolution of the diametric compression of 𝜎1 − 𝜎2 = = , (21)
𝜋𝑡𝑟 𝑡
a circular disc [48], the equations describing the stress components can
and:
be written as:
4𝑃
⎧ ⎫ 𝑓 = . (22)
𝜋𝑟𝑁
2𝑃 ⎪ 𝑥2 (𝑟 + 𝑦) 𝑥2 (𝑟 − 𝑦) ⎪ 𝑃
𝜎𝑥 = −
𝜋𝑡 ⎪ [𝑥2 + (𝑟 + 𝑦)2 ]2 [𝑥2 + (𝑟 − 𝑦)2 ]2 ⎬
⎨ + + As shown in Fig. 13, three stages, namely, P = 1 kN, 2 kN, 3 kN, re-
⎪ 𝜋𝑟𝑡
⎩ ⎭ spectively, were selected. The fringe orders at the center point were
calculated by the proposed method, the fringe values were determined
⎧ ⎫
2𝑃 ⎪ (𝑟 + 𝑦)3 (𝑟 − 𝑦)3 ⎪ 𝑃 using Eq. (22) (the results are listed in Table 4), and the average of the
𝜎𝑦 = − ,
𝜋𝑡 ⎨ 2⎬
[ ]2 + + (19) three fringe values was selected as the fringe value for all stages.
⎪ 𝑥 + (𝑟 + 𝑦)
2 2
[𝑥 + (𝑟 − 𝑦) ] ⎪
2 2 𝜋𝑟𝑡
⎩ ⎭ The whole field principal stress difference of the three stages was
⎧ ⎫ calculated to show the stress changes during the loading process. Theo-
2𝑃 ⎪ 𝑥(𝑟 + 𝑦)2 𝑥(𝑟 − 𝑦)2 ⎪ retically, values for all stages in the whole continuous loading progress
𝜏𝑥𝑦
𝜋𝑡 ⎨ ]2 ⎬
=− [ ]2 + [ can be calculated, except for the plastic regions. To verify the accuracy
⎪ 𝑥 + (𝑟 + 𝑦)
2 2
𝑥 + (𝑟 − 𝑦) ⎪
2 2
⎩ ⎭ of the experimental results, the theoretical results of the principal stress
difference calculations from Eq. (20) for the same loading pressure were
where r is the disc radius, t is the thickness of the circular disc, and
also calculated, as shown in Fig. 14.
P is the loading pressure. Theoretically, the information shown in
The calculated results of the experiments and theory match well
the photoelasticity fringe patterns is the principal stress difference, so
across almost the entire disc, except for the loading pressure points,
Eq. (19) can be rewritten as:
( ( )) as shown in Fig. 14. Furthermore, the calculated results of the experi-
√ 4𝑃 𝑟2 − 𝑥2 + 𝑦2 mental and theoretical principal stress differences in the horizontal and
𝜎1 − 𝜎2 = (𝜎𝑥 − 𝜎𝑦 )2 + 4𝜏𝑥𝑦
2 = (( ) ), (20)
𝜋𝑡𝑟 𝑥2 + 𝑦2 + 𝑟2 − 4𝑦𝑟2 vertical directions of the different stages (see Fig. 14(a–g)) were counted

255
Y. Ju et al. Optics and Lasers in Engineering 100 (2018) 248–258

Fig. 17. Influence of the initial light intensity: (a) distribution of the initial light intensity, (b) errors caused by the initial light intensity, (c) first frame used to calculate the whole-process
fringe orders, and (d) the error regions caused by the non-uniformity of the initial light intensity.

by the curve to x-axis or y-axis (see Fig. 15(a) and (b)), where the solid a certain point in the loading diameter can be expressed as:
lines indicate the experimental results and the dotted lines show the 𝑦
theoretical results. It is evident that the experimental and theoretical 𝑑=2 (𝜎𝑦 − 𝜇𝜎𝑥 )𝑑𝑦, (24)
lines at the same stage are close to each other, which indicates that this ∫
0
method can accurately determine the whole-field principal stress differ-
ence. However, when there are too many fringes in the loading pres- then 𝜎 x and 𝜎 y can be substituted into Eq. (24), and the results of the
sure points, the stress is so high that plastic zones form, which results in calculation can be written as:
( ( ))
fringes that cannot be clearly differentiated. Consequently, these regions 2𝑃 𝑟+𝑦
𝑑= (1 − 𝜇)𝑦 − 2𝑟𝑙𝑛 , (25)
in the curve of the principal stress difference in the vertical direction to- 𝜋𝑟𝑡𝐸 𝑟−𝑦
ward the y-axis were removed, as shown in Fig. 14(b) where the Y-axis where E is the elastic modulus and 𝜇 is Poisson’s ratio. According to
extends from y = −22 to y = 22. In addition, the X-axis in Fig. 14(a) ex- [14], E = 3.1 GPa and 𝜇 = 0.38. Hence, the displacement of the stage for
tends from x = −24 to x = 24 due to the influence of the boundary. In P = 3 kN is:
summary, the experimental results are close to the theoretical results, ( ( ))
25 + 𝑦
with the exception of some localized regions. 𝑑 = 0.0041 0.62𝑦 − 50𝑙𝑛 . (26)
25 − 𝑦
To illustrate the influence of the deformation on the accuracy of
the calculation results, the fringe orders of some special points in the
3.2. Inaccuracy analysis loaded diameter (see Figs. 16(a) and (b)) can be determined by identi-
fying the darkest and brightest points in the dark field of the circular
3.2.1. The influence of deformation polariscope. According to Eq. (10), the darkest points are the integer
Based on Eq. (19), in the loading diameter of the line at x = 0, the fringe orders, and the brightest points are the half-integer fringe orders.
stress components can be written as: Theoretically, the results obtained by this method are accurate, so the
absolute value of the fringe orders difference of the two methods can be
𝑃
𝜎𝑥 = regarded as a parameter representing the influence of deformation, as
𝜋𝑟𝑡
( ) shown in Fig. 16(a) and (b). In the figure, the horizontal axis represents
𝑃 4𝑟2 . (23) the displacement of the points for which the fringe orders were calcu-
𝜎𝑦 = 1−
𝜋𝑟𝑡 𝑟2 − 𝑦2 lated by the two methods, and the vertical axis shows the absolute value
𝜏𝑥𝑦 = 0 of the fringe orders difference of the two methods.
Figs 16(a) and (b) indicate that the change of fringe orders differ-
The state of the plane stress is assumed for the diametric compression ence can be divided into two stages: (1) when the displacement is less
of the circular disc and, according to Hooke’s law, the displacement of than 0.4 mm, the fringe orders difference in most counted points does

256
Y. Ju et al. Optics and Lasers in Engineering 100 (2018) 248–258

not exceed 0.1, and the rate with the increment of displacement is close points led to an emerging zone of plasticity where the stress-optics
to zero; (2) when the displacement is greater than 0.4 mm, the fringe or- laws do not apply.
ders difference maintains sustained and rapid growth. The major reason (3) This method can be used to simplify the experimental operation over
for this phenomenon is that when the displacement increases, the fringe the entire process. The experiments were carried out on a single
intensity in these regions is very high, and the fringe orders change along model of circular polariscope, which eliminates the need to continu-
with the increment of displacement; hence, the displacement will signif- ally adjust the camera angles. Moreover, the independent calculation
icantly influence the accuracy of the calculated results. In other words, of the fringe order based on the light intensity changes of each point
the accuracy of the calculated results drops quickly when the fringe in- in the loading process means that the accumulated errors from the
tensity exceeds a certain threshold, and these regions will reach a plas- inaccurate results of other points will be eliminated. Meanwhile, the
tic state where the photoelasticity stress law is not applicable. These determination of fringe orders mainly relies on the resolution ratio
regions are not considered in this paper; hence, the error in the calcu- of the frames, so dense high-order fringes can be calculated as long
lated results in most regions of the disc is acceptable because the largest as the fringes are clearly discernible. In addition, the fringe orders in
difference divided by the fringe orders calculated by second method is the elastic stage in the continuous loading process can be determined
less than 2.5%. Note that this method should only be applied within in a way that is favorable for visualizing the dynamic evolution of
elastic deformation areas where the photoelastic stress law is applica- the whole-field stress.
ble and fringes can be clearly identified. It is extremely challenging for
the method to extract and analyze fringes in areas where plastic defor- In conclusion, this method is well suited for determining the whole-
field stress, which will contribute to the further development of digi-
mation takes place because high-density fringes usually occur in these
tal photoelasticity methods. Especially because the calculation of whole
areas, and these are difficult to distinguish and extract. Thus, the photoe-
lastic stress law may not be applicable for determining the relationship field stress is independent on the complex structure of models, this
method has potential to calculate the fringe orders of these models,
between the stresses and fringe orders of models.
which will be studied in the next work. This paper presents the pre-
3.2.2. Influence of the initial light intensity distribution liminary results of our research. It is noteworthy that our method was
Based on Section 2.3.1, the integer part of the fringe orders is deter- established based on the regular changes in the light intensity at the pix-
mined by the point number, as shown in Fig. 5. However, the hetero- els when continuous loads are applied; however, the proposed method
geneity of the initial light intensity distribution, i.e., the light intensity can only be applied to photoelastic fringes that can be clearly identified
distribution of the initial frame in Fig. 17(a), will lead to the counted using high-resolution digital imaging techniques. In addition, it is dif-
point number in Fig. 5 being higher than the integer part of the actual ficult for the method to extract and analyze the fringes in areas where
fringe orders. For example, Fig. 17(b) shows the change curve of light plastic deformation takes place. In the future, we intend to work toward
intensity of a certain point when the load was so small that the first solutions for the problems identified in this paper, such as regions where
fringe had not yet formed, but the points number used to calculate the the fringe density is too high to distinguish individual fringes, although
fringe orders indicated the first fringe order had formed. To show the we recognize that this will require continued development of the me-
influence across the whole field, the integer part of the fringe orders of chanics of photoplasticity and the resolution of digital cameras. We also
the 32nd frame (see Fig. 17(c)) was calculated, as shown in Fig. 17(d), noted that the emergence of unloading in some areas will reduce the
which indicates the fringe orders reached one or more, as represented fringe orders, which will influence the accuracy of the calculated re-
by the white regions marked by red ellipses in Fig. 17(d), before the sults. However, this happens when the load is so large that plastic areas
real first order fringe formed. Thus, to eliminate the influence of the or cracks form, and rarely emerges in the elastic stage. This behavior
initial light intensity, the first frame selected to calculate the fringe or- will be explored in a future work.
ders should not be the first frame before loading, but the frame when
the influence of the initial light intensity is small enough to be ignored. Acknowledgements
It should be noted that the influence of the initial light intensity will
diminish or disappear as the light intensity formed by the birefringence The authors gratefully acknowledge the financial support received
gradually gains a dominant position in the increment of circular disc from the National Natural Science Foundation of China [Grant nos.
deformation. 51374213 and 51674251], the State Key Research Development Pro-
gram of China [Grant no. 2016YFC0600705], the National Natural Sci-
4. Conclusion ence Foundation for the Distinguished Young Scholars of China [Grant
no. 51125017], the Science Fund for Creative Research Groups of the
In this paper, we proposed a novel method to determine the whole National Natural Science Foundation of China [Grant no. 51421003],
field fringe orders. With this method, the fringe order of each point on the Fund for Creative Research and Development Group Program of
the model is independently determined. The theoretical background of Jiangsu Province [Grant no. 201427], and the Priority Academic Pro-
this method is based on our fundamental optics knowledge and corre- gram Development of Jiangsu Higher Education Institutions [Grant no.
sponding experiments related to the diametric compression of a circular PAPD2014].
disc that were carried out to verify the efficacy of the proposed method.
Our conclusions can be summarized as follows: References
(1) A 3D printer with photoelastic filament was used to prepare trans- [1] Zhou XG, Cao CJ, Yuan JY. The research actuality and major progresses on the quan-
parent models for capturing and visualizing continuous whole-field titative forecast of reservoir fractures and hydrocarbon migration law. Adv Earth Sci
stress using photoelasticity testing. The method of using a circular 2003;18:398–404.
[2] Zhang LK, Wang ZL, Qu ZH. Physical simulation experiment of gas migration in
transparent disc in diametric compression was used to introduce the
sandstone porous media. Acta Geol Sin 2007;81:539–44.
proposed method. The results of testing indicate the method works [3] Hou Q, Li H, Fan J, Ju Y, Wang T, Li X, et al. Structure and coalbed methane occur-
properly, and can be used to determine the whole field fringe orders. rence in tectonically deformed coals. Sci China Earth Sci 2012;55:1755–63.
[4] White CM, Smith DH, Jones KL, Goodman AL, Jikich SA, LaCount RB, et al. Seques-
(2) A comparison between the theoretical and experimental results for
tration of carbon dioxide in coal with enhanced coalbed methane recovery a review.
different loads showed a good match, except for the loading points. Energy Fuels 2005;19:559–724.
On one hand, the influence of the stress concentration in these re- [5] Wang J. On area-specific underground research laboratory for geological disposal of
gions was obvious and resulted in the formation of high-density high- level radioactive waste in China. JRMGE 2014;6:99–104.
[6] Shaik AR, Rahman SS, Tran NH, Tran T. Numerical simulation of fluid-rock cou-
fringes, which could not be captured by a conventional digital cam- pling heat transfer in naturally fractured geothermal system. Appl Therm Eng.
era. On the other hand, the high stress concentration at the loading 2011;31:1600–6.

257
Y. Ju et al. Optics and Lasers in Engineering 100 (2018) 248–258

[7] Kim SO, Kim JJ, Yun ST, Kim KW. Numerical and experimental studies on cad- [27] Yao XF, Jian LH, Xu W, Jin GC, Yeh HY. Digital shifting photoelasticity with opti-
mium (II) transport in kaolinite clay under electrical fields. Water Air Soil Pollut cal enlarged unwrapping technology for local stress measurement. Opt Laser Tech
2003;150:135–62. 2005;37:582–9.
[8] Tarpani JR, Ruckert CO, Milan MT, Silva RV, Rosato A, Pereira RN, et al. Estimating [28] Baek TH, Kim MS, Hong DP. Fringe analysis for photoelasticity using image process-
fatigue life under variable amplitude loading through quantitative fractography – a ing techniques. Int J Soft Eng Appl 2014;8:91–102.
case study. Eng Fail Anal 2004;11:547–59. [29] Prasad VS, Madhu KR, Ramesh K. Towards effective phase unwrapping in digital
[9] Morgeneyer TF, Starink MJ, Sinclair I. Evolution of voids during ductile crack prop- photoelasticity. Opt Lasers Eng 2004;42:421–36.
agation in an aluminium alloy sheet toughness test studied by synchrotron radiation [30] Ramji M, Ramesh K. Whole field evaluation of stress components in digital photoe-
computed tomography. Acta Mater 2008;56:1671–9. lasticity —issues, implementation and application. Opt Lasers Eng 2008;46:257–71.
[10] Lempp C, Shams KM, Jahr N. Approaches to stress monitoring in deep boreholes for [31] Lei Z, Yun H, Zhao Y, Xing Y, Pan X. Study of shear transfer in Al/epoxy joint by
future CCS projects. Environ Earth Sci 2012;67:435–45. digital photoelasticity. Opt Lasers Eng 2009;47:701–7.
[11] Mazaira A, Konicek P. Intense rockburst impacts in deep underground construction [32] Ramakrishnan V, Ramesh K. Scanning schemes in white light photoelasticity – part
and their prevention. Can Geotech J 2015;52:150421143936002. I: critical assessment of existing schemes. Opt Lasers Eng 2016;92:129–40.
[12] Bai M, Roegiers JC, Elsworth D. Poromechanical response of fractured-porous rock [33] Ramakrishnan V, Ramesh K. Scanning schemes in white light photoelasticity – part
masses. JPSE 1995;13:155–68. II: novel fringe resolution guided scanning scheme. Opt Lasers Eng 2016;92:141–9.
[13] Sun B, Zhu Z, Shi C, Luo Z. Dynamic mechanical behavior and fatigue damage evo- [34] Ajovalasit A, Petrucci G, Scafidi M. Review of RGB photoelasticity. Opt Lasers Eng
lution of sandstone under cyclic loading. Int J Rock Mech Min Sci 2017;94:82–9. 2015;68:58–73.
[14] Makhnenko RY, Harvieux J, Labuz JF. Paul‐Mohr‐Coulomb failure surface of rock [35] Quiroga JA, Servin M, Marroquin JL. Regularized phase tracking technique for
in the brittle regime. Geophys Res Lett 2015;42:6975–81. demodulation of isochromatics from a single tricolor image. Meas Sci Technol
[15] Dimas LS, Giesa T, Buehler MJ. Coupled continuum and discrete analysis of 2002;13:132–40.
random heterogeneous materials: elasticity and fracture. J Mech Phys Solids [36] Quiroga JA, García-Botella A, Gómez-Pedrero JA. Improved method for isochromatic
2014;63:481–90. demodulation by RGB calibration. Appl Opt 2002;41:3461–8.
[16] Farhat C, Avery P, Chapman T, Cortial J. Dimensional reduction of nonlinear finite [37] Ajovalasit A, Petrucci G, Scafidi M. RGB photoelasticity: review and improvements.
element dynamic models with finite rotations and energy‐based mesh sampling and Strain 2010;46:137–47.
weighting for computational efficiency. Int J Numer Method Eng 2014;98:625–62. [38] Grewal GS, Dubey VN, Claremont DJ. Isochromatic demodulation by fringe scan-
[17] Wang SY, Sloan SW, Sheng DC, Yang SQ, Tang CA. Numerical study of failure be- ning. Strain 2010;42:273–81.
haviour of pre-cracked rock specimens under conventional triaxial compression. Int [39] Jones IA, Wang P. Complete fringe order determination in digital photoelasticity
J Solids Struct 2014;51:1132–48. using fringe combination matching. Strain 2010;39:121–30.
[18] Forte P, Paoli A, Razionale AV. A CAE approach for the stress analysis of gear models [40] Ajovalasit A, Petrucci G. Developments in RGB photoelasticity. Appl Mech Mater
by 3D digital photoelasticity. IJIDM 2015;9:31–43. 2005;3–4:205–10.
[19] Rosakis AJ, Samudrala O, Coker D. Cracks faster than the shear wave speed. Science [41] Ng TW. Photoelastic stress analysis using an object step-loading method. Exp Mech
1999;284:1337–40. 1997;37:137–41.
[20] Takacs IA, Botean AI, Hardau M, Chindris S. Displacement-stress distribution in a [42] Ng TW, Chau FS. Automated analysis in digital speckle shearing interferometry using
femoral bone by optical methods. Procedia Tech 2015;19:901–8. an object step-loading method. Opt Commun 1994;108:214–18.
[21] Ramesh K, Hariprasad MP, Bhuvanewari S. Digital photoelastic analysis applied to [43] Zuccarello B, Tripoli G. Photoelastic stress pattern analysis using Fourier trans-
implant dentistry. Opt Lasers Eng 2016;87:204–13. form with carrier fringes: influence of quarter-wave plate error. Opt Lasers Eng
[22] Ju Y, Xie H, Zheng Z, Lu J, Mao L, Gao F, et al. Visualization of the complex 2002;37:401–16.
structure and stress field inside rock by means of 3D printing technology. Sci Bull [44] Zhang M, Shimada H, Sasaoka T, Matsui K, Dou L. Evolution and effect of the stress
2014;59:5354–65. concentration and rock failure in the deep multi-seam coal mining. Environ Earth
[23] Ju Y, Wang L, Xie H, Ma G, Zheng Z, Mao L. Visualization and transparentization Sci 2014;72:629–43.
of the structure and stress field of aggregated geomaterials through 3D printing and [45] Fan Y, Lu W, Zhou Y, Yan P, Leng Z, Chen M. Influence of tunneling methods on
photoelastic techniques. Rock Mech Rock Eng 2017;50:1383–407. the strain- burst characteristics during the excavation of deep rock masses. Eng Geol
[24] Ju Y, Zheng Z, Xie H, Lu J, Wang L, He K. Experimental visualisation methods for 2015;201:85–95.
three- dimensional stress fields of porous solids. Exp Tech 2017;41:331–44. [46] Busetti S, Mish K, Hennings P, Reches ZE. Damage and plastic deformation of reser-
[25] Zhenkun L, Wei Q, Libo D. Flexural effects of sandwich beam with a plate insert voir rocks: part 2. Propagation of a hydraulic fracture. AAPG Bull 2012;96:1711–32.
under in-plane bending. Opt Laser Tech 2012;44:1223–31. [47] Theocaris BPS, Gdoutos EE. Matrix theory of photoelasticity. Springer-Verlag; 1979.
[26] Kramer SL, Beiermann BA, White SR, Sottos NR. Simultaneous observation of [48] Jianhong Y, Wu FQ, Sun JZ. Estimation of the tensile elastic modulus using Brazilian
phase- stepped images for photoelasticity using diffraction gratings. Exp Mech disc by applying diametrically opposed concentrated loads. Int J Rock Mech Min Sci
2013;53:1343–55. 2009;46:568–76.

258