Journal of Nuclear Materials 503 (2018) 205e225

Contents lists available at ScienceDirect

Journal of Nuclear Materials

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

Experimental and computational correlation of fracture parameters

KIc, JIc, and GIc for unimodular and bimodular graphite components
Awani Bhushan, S.K. Panda*
Department of Mechanical Engineering, Indian Institute of Technology (Banaras Hindu University), Varanasi 221005, India

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

Article history: The influence of bimodularity (different stress ~ strain behaviour in tension and compression) on fracture
Received 7 November 2017 behaviour of graphite specimens has been studied with fracture toughness (KIc), critical J-integral (JIc) and
Received in revised form critical strain energy release rate (GIc) as the characterizing parameter. Bimodularity index (ratio of
5 March 2018
tensile Young's modulus to compression Young's modulus) of graphite specimens has been obtained
Accepted 5 March 2018
from the normalized test data of tensile and compression experimentation. Single edge notch bend
Available online 9 March 2018
(SENB) testing of pre-cracked specimens from the same lot have been carried out as per ASTM standard
D7779-11 to determine the peak load and critical fracture parameters KIc, GIc and JIc using digital image
Keywords:
3D J integral
correlation technology of crack opening displacements. Weibull weakest link theory has been used to
Bimodularity evaluate the mean peak load, Weibull modulus and goodness of fit employing two parameter least
Contour integral square method (LIN2), biased (MLE2-B) and unbiased (MLE2-U) maximum likelihood estimator. The
Fracture toughness stress dependent elasticity problem of three-dimensional crack progression behaviour for the bimodular
Graphite graphite components has been solved as an iterative finite element procedure. The crack characterizing
Strain energy release rate parameters critical stress intensity factor and critical strain energy release rate have been estimated with
Weibull the help of Weibull distribution plot between peak loads versus cumulative probability of failure.
Experimental and Computational fracture parameters have been compared qualitatively to describe the
significance of bimodularity. The bimodular influence on fracture behaviour of SENB graphite has been
reflected on the experimental evaluation of GIc values only, which has been found to be different from the
calculated JIc values. Numerical evaluation of bimodular 3D J-integral value is found to be close to the GIc
value whereas the unimodular 3D J-value is nearer to the JIc value. The significant difference between the
unimodular JIc and bimodular GIc indicates that GIc should be considered as the standard fracture
parameter for bimodular brittle specimens.
© 2018 Elsevier B.V. All rights reserved.

1. Introduction only anisotropic and orthotropic materials such as composites, but

Bimodularity introduces uncertainty in failure characteristics of
many structures leading to unreasonable compromise on factor of
safety of high risk structures. The phenomena though is present at
elemental scale for most materials, the scale of their severity varies
from material to material, being predominant for brittle materials,
where a safe guard of flow stress induced plasticity is not present.
Therefore, many catastrophic fracture of brittle structures remain
unanswered leading to critical design fallacy.
Structural materials exhibiting different stress-strain curves in
compression and tension are termed as bimodulus materials. Not
* Corresponding author.
E-mail address: pandask@bhu.ac.in (S.K. Panda).
also some traditional isotropic materials as ceramics, graphites may
also have different moduli in tension and compression. In the case
of bimodulus materials the constitutive matrix is a function of
stress. Though the stress-strain relationship is actually curvilinear,
but it is approximated as bilinear with different slopes. Hence the
analysis of structures made up from bimodulus materials is more
involved. Two basic material models viz. Ambartsumyan [1e7] and
Bert [8e11] are being most widely used for characterizing such
bimodulus behaviour. Ambartsumyan material model is based on
the criterion of positive-negative signs of principal stress state at a
point in a stressed body. This model has found its application
mostly to isotropic materials having bimodulus characteristic. Bert
material model is based on the criterion of positive-negative signs
in the longitudinal strain of fibers in orthotropic materials, and

https://doi.org/10.1016/j.jnucmat.2018.03.012
0022-3115/© 2018 Elsevier B.V. All rights reserved.
206 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225
hence has its significance in research for laminated composites. The
elasticity theory of Ambartsumyan material model considers the
dependence of moduli both on material properties and also the
state of stress at the point. In retrospect, elastic modulus has
nonlinear characteristics and it is related to the material, shape,
boundary conditions, and external loads on the structure. The main
difficulty in the analysis of bimodulus planar structures is that of
locating the neutral surface. Since the element moduli depend on
the sign of the stress which is unknown a priori, the iterative
techniques need to be developed to find the layer in which there is
no strain. Along with this, a simplified linearized mechanical model
is necessary for optimization study of such problems. In this respect
Finite Element Method (FEM) based analytical and iterative pro-
cedures are found to be appropriate and most comprehensive for
studying the reliability and failure characteristics of bimodulus
material structures. The bimodular Ambartsumyan linear model
based on the rules of elastic continuum mechanics has the
assumption that the stressed body is continuous, homogeneous,
and isotropic and all deformations are small enough to have theory
of superposition valid for all response characteristics. The linear-
ized constitutive model simplifies the tension-compression stress
strain plots into two straight lines, whose tangents at the origin are
discontinuous.
In this work an endeavor has been made to systematically
analyze the fracture behaviour of bimodular material graphite be-
ing chosen for the study. The choice of graphite is because, apart
from being a reliable refractory material having high temperature
stability, its applications are manifold in critical areas of nuclear
energy.
Graphite is currently being used for the construction of the
major core components such as the fuel block, reflector, moderator
and core support critical structure in a nuclear reactor. This will be a
key material in the development of Very High Temperature Reactor
(VHTR) for the six next-generation nuclear reactor systems within
the Generation IV International Forum [12]. The VHTR is considered
to be the Next Generation Nuclear Reactor (NGNR) in the evolu-
tionary development of high-temperature gas-cooled reactors with
significant advantage of inherent safety, high thermal efficiency,
process heat application capability, low operation and maintenance
costs, and modular construction [13,14].
Graphite material characterization can have widely different
textures and pore-size distributions, as well as the subcritical crack
like formations. The physical properties of artificial graphite are
mostly anisotropic due to the hexagonal layer lattice structure of its
crystallites and its degree of orientation during processing [15e17].
As such different grades of graphites can be manufactured with
different average grain sizes. Typically, coarse-grained material has
grains larger than 4 mm; medium-grained material has grains
smaller than 4 mm; fine-grained material has grains smaller than
100 mm; and superfine, ultrafine, and micro-fine materials have
grain sizes smaller than 50 mm, 10 mm, and 2 mm, respectively. Nu-
clear grade graphite has grain size falls within a range from me-
dium to ultrafine. As graphites are manufactured by extrusion or
molding process, the resulting grain structure has a biased orien-
tation with respect to material coordinates. Usually, material
properties are measured either along the grain orientation i.e.
parallel to the extrusion direction and perpendicular to the molding
axis; or against the grain orientation i.e. perpendicular to the
extrusion direction and parallel to the molding axis. This explains
the anisotropic or more accurately the transverse isotropic behav-
iour of graphite materials [18,19]. Due to large scatter in flaws and
its distribution, size scaling and extrapolation from lab scale
coupon tests to actual components in reactor service has been
difficult for graphite components. Therefore, proper design of test
specimen configuration, test method, and the analysis of
mechanical properties test data for extrapolation should be carried
out for a certain degree of reliability and satisfactory performance
of such structures. In this regard, Weibull statistical methods are
found to be efficient in not only reducing the number of experi-
mentations for predicting the mechanical behaviour, but also based
on this ASTM standard procedures and analytical expressions are
developed for size scaling from component to component and from
one loading configuration to another. The probability of fracture
from material test results of lab scale specimen, translation and
scaling of such data to predict the probability of fracture of actual in
service component involve a need to understand the influence of
the state of stress on the fracture strength of test coupons and
components [20,21].
When a material exhibits bimodular characteristics, this prob-
lem is challenging, because the characterization of state of stress
invokes quantification of elastic moduli whether in tension or
compression. In this work, the graphite is being considered to be
isotropic exhibiting bimodularity. Within the nuclear reactor
environment, graphite components are subjected to neutron irra-
diation under high temperature, resulting in changes in physical
and microstructural properties. Failure and fracture behaviour of
graphite components are very critical to the reliability of core
structures. Presence of crack like defects drastically enhances the
probability of failure of graphite components. Material parameters
such as elastic modulus and fracture resistance happen to be the
lead parameters in postulating the life of such structures. Therefore,
accurate characterization of the fracture properties of nuclear
graphite qualitatively and quantitatively is important to the integ-
rity or safety of a VHTR [13]. Apart from critical fracture parameters
such as stress intensity factor, strain energy release rate and critical
value of J-integral, the fracture behaviour of nuclear grade graphite
material is also dependent on the size and shape of specimen
[22e25]. In flexural condition, probability of failure is affected by
the state of stress of tensile or compressive regions. Then evaluating
the location neutral plane is significant to understand the criticality
of stress state. When a material exhibit different elastic character-
istic in tension and compression, then accurate designing of such
structures poses multiple challenges of stress dependent elasticity
problem. The singularity of crack tip stress field under such loading
state become complicated to find a closed form solution. Ignoring
such real scale phenomenon leads to unacceptable failure evi-
dences of bimodular material structures. Therefore, modified
design procedures are to be laid to take into account the validity of
different stress vs strain behaviour. Materials exhibiting different
mechanical behaviour in tension and compression are addressed in
literature as bimodular material. The bimodularity has been found
to severely affect the probability of failure of cylindrical specimens
tested in three point and four point flexural loading conditions [26].
The concept of Bimodular material was first explored by Saint-
Venant [27]and it was revisited by Timoshenko [28]. The concept
of bimodulus material was extended to two-dimensional materials
by Ambartsumyan [2e7,29,30] postulating a Multi-modulus elas-
ticity theory. He suggested a three-dimensional, stress-strain law
for isotropic bimodulus materials and developed such relations for
orthotropic plates. Tabaddor [31e33] developed constitutive
equations for the more general case of bimodulus materials
formulating the generalized stress-strain laws for the anisotropic
bimodularity. He employed finite element procedures for analyzing
the two dimensional bimodulus beam. N. Kamiya presented a
method of analysis of the bending of bimodulus elastic plates
employing Ambartsumyan-Khachatryan's model for isotropic
bimodulus materials [34,35]. This problem may be reduced to the
conventional problem of minimizing the potential energy of the
plate as a whole. A simply supported thin square plate subjected to
lateral load was analyzed numerically by simplex method. Results
 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225
of the calculation show that the effect of the difference between the
tensile and compressive elastic moduli on the deformation of the
plate may be substantial. Jones material model based on the prin-
ciples of anisotropic elasticity was developed for bimodular mate-
rials [36e38]. Weibull probabilistic failure analysis and fracture
mechanics procedures were used to study the failure characteris-
tics of nuclear grade graphite [39,40]. Two and three parameter
Weibull models have been compared for their respective failure
distributions.
Ouagne et al. [41] used R-curve analysis to study the crack
growth behaviour of polygranular graphite. Field strain mapping
technique was used to show the development of the distribution of
localized deformation, which can be linked to the microstructure
and damage processes in nuclear graphite [42]. Weibull parameter
estimation of the strength distribution was carried out to provide
the detail information regarding the survival probability or the
reliability under stress of the nuclear grade graphite using me-
chanical properties as specified in the ASTM specification D7219-14
[43]. Karthik et al. [44] studied the damage process in nuclear grade
graphite by in situ transmission electron microscopy and observed
the atomic level processes involved in the swelling and crack-
closing in nuclear grade graphite under electron irradiation. Dou-
ble torsion (DT) technique was adapted to describe stable crack
propagation and calculating the J-integral from in-plane displace-
ment fields [45]. Rising R-curve behaviour was observed, which is
endorsed to the formation of the fracture process zone, while crack
bridging and distributed micro-cracks were responsible for the
increase in fracture resistance. Hindley et al. [46,47] employed
statistical distribution to represent the experimental material
strength of NBG-18 nuclear graphite and studied the probability of
failure of graphite core components. A relevant statistical fit is
determined and the goodness of fit was also evaluated for each data
set. However, the literature studies indicate that bimodularity be-
ing a state of the art phenomenon has not been pursued for the
computational stochastic study or experimental procedures. In the
absence of closed form analytical solutions for fracture study of
bimodularity, it becomes natural to give impetus to the mentioned
procedures to characterize the failure modes of high risk bimodular
material structures. The present work constitutes experimentation;
finite element method based computational fracture mechanics
procedures and Weibull stochastic modeling principles to quantify
crack propagation behaviour of bimodular graphite with strain
energy release rate as the fracture parameter. Two large set of
samples are tested in tension and compression to evaluate the
change in stress vs strain behaviour and determine the elastic
modulus ratio of tension to compression.
The data sets are also normalized for ease of comparison, and
combined into one representative data set. The validity of this
approach has been demonstrated in subsequent sections. Modulus
of elasticity for tension and compression has been obtained
experimentally by taking the average values for defining the
bimodular ratio for the graphite component. The same log of
graphite is used to build the fracture specimen using wire cut EDM
machines. The specimens are tested to estimate the critical stress
intensity factor at peak load obtained in the load versus displace-
ment plot. For each data point on load-displacement plot, energy
release rate has been calculated. The critical energy release rate has
been determined from the energy release rate and corresponding
crack extension plot. The critical value of the path independent J-
integral is obtained neglecting the plastic zone for brittle materials
[48]. Thereafter computational studies of stress dependent elas-
ticity problem have been carried out to signify the appropriateness
of bimodulus elasticity for characterizing the fracture parameter.
The influence of bimodularity on the fracture parameter has been
demonstrated with the help of experiments, statistical analysis, and
207
computational numerical model analysis. The following sections
elaborate on the laid down procedures and observations.
2. Experiments
Experimentation for tensile, compressive and fracture test data
have been conducted as per relevant ASTM standards. The fracture
data has been obtained by using digitalized image correlation
software with continuous image capturing through high speed
camera attachment to the INSTRON machine set up.
2.1. Specimen preparation
The tensile and compressive specimens are prepared from the
cuboidal shape graphite log of size as 1040
650
350 mm3 and
cut down in the way that it occupies horizontal orientation or
against the gravity as shown in Fig. 1. The modulus of elasticity in
tension and compression has been evaluated for this sample of
horizontal oriented specimens. The fracture specimens are also
prepared in the similar orientation according to ASTM D7779
standard [49]. Fracture experimentation have been conducted to
determine the critical stress intensity factor and strain energy
release rate for the three point single edged notched bend (SENB)
specimens prepared from the same log and subjected to displace-
ment controlled point load.
2.2. Tensile and compressive test
High purity machined graphite tensile and compressive speci-
mens of 20 mm gage diameter have been tested as per ASTM C-749
and ASTM C-695 for assessing the bimodular behaviour. The gauge
length of tensile specimen is 80 mm and for compression specimen
is 40 mm [50,51]. Proper care and specimen protection has been
invoked during the testing for determining appropriate cross head
speed so that the specimen should not break at the fixture clamp
during tensile loading. Twenty specimens from each category are
tested at room temperature using INSTRON UTM and the results are
recorded for assessing the bimodular behaviour. Utmost care has
been taken for preparing the specimens as per ASTM standard with
reference to surface finish and planeness of the specimens within
the tolerance values. A crosshead speed of 0.2 mm/min has been
maintained throughout the testing.
2.3. Three point single edge notch bend (SENB) test
The fracture specimens have been prepared from same graphite
log for which the tensile and compressive specimens have been
made. Qualitative validation of the effect of bimodularity has been
carried out experimentally for single edge three-point bend
notched graphite specimens. The bimodular graphite specimens
are tested for evaluating fracture toughness as per ASTM D7779
[49]. The SENB specimens have notch length and tip radius as 5 mm
and 0.1 mm respectively. The span length of the specimen is 50 mm
having square cross-section (10 mm
10 mm) as displayed in
Fig. 2. Thirty specimens are prepared to make an effective conclu-
sion over the influence of bimodularity on the fracture character-
istics. Wire cut electro discharge machining has been used to
develop the notch with 300 notch angle at the tip. The specimens
are tested with specific cross head speed of 0.01 mm/min for
obtaining a smooth load displacement plot. Crack opening
displacement and load line displacements are correlated using
digital image correlation software arrangements.
All mechanical testings have been carried out in a 64 bit 30 KN
INSTRON UTM attached with Blue hill 3 software and a 5 KN load
cell for the three-point bend testing. The load measurement
208 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225
Fig. 1. The specimens making in the two orientations vertical (with gravity:Y-axis) and horizontal (against gravity: X-axis) within specific graphite block size
1040
650
350 mm3.
accuracy is ±0.5% of reading down to 1/100 of load cell capacity. The
load cell accuracy is ±0.25% of the indicated force. The machine has
been calibrated the strain measurement accuracy is ±0.5% of
reading down to 1/50 of full scale and the crosshead speed accuracy
is ±0.2% of the set speed. The cross head speed chosen for the SENB
testing has been 0.01 mm/min. The deformation measurement for
the testing between two consecutive intervals of increasing force is
close to 5
10 5 mm, which might give an insight to the mea-
surement accuracy of the machine.
2.3.1. Evaluation of fracture toughness and strain energy release
rate
The critical fracture parameter for steady state crack propaga-
tion has been estimated from the peak load value obtained in the
load displacement plot. Fracture toughness KIc of the graphite
material has been calculated using ASTM standard D7779-11 [49].
! pffiffiffiffiffiffiffiffiffiffiffi ! determining material fracture toughness. The mean peak load has
Pmax S:106 3 a=W
KIC ¼g
BW 3=2 2ð1  a=WÞ3=2
where, g ¼ gða=WÞ
g ¼ A0 þ A1 ða=WÞ þ A2 ða=WÞ2 þ A3 ða=WÞ3 þ A4 ða=WÞ4
þ A5 ða=WÞ5
‘Pmax’ is the peak load, ‘S’ is the support span length, ‘a’ is the
initial crack length, ‘B’ is the width and ‘W’ is the thickness of the
SENB specimen. The value of coefficients for the above polynomials
(A0, A1, A2, A3, A4 and A5) have been taken as per ASTM D7779-11
[49]. The ASTM procedure for evaluation of fracture toughness
and energy release rate are different for the graphite specimen. The
critical stress intensity factor depends on the peak load obtained in
the experiments and geometrical parameters. The test method for
calculating strain energy release rate with respect to progressive
crack extension requires plotting of load versus load line
displacement of individual data points. Then, the critical energy
release rate is obtained by normalizing the initial non-linearity
with a linear relationship.
3. Results and discussion
Thirty three point bend notched graphite specimens are tested
to evaluate the fracture toughness of the material. Weibull statis-
tical model has been developed to evaluate the mean peak load for
been employed in the finite element simulations to characterize the
(1)
influence of bimodularity on the fracture behaviour of the graphite
components. It is important that working on fracture specimens,
the machine vibration should be minimized for accuracy of the
results. Even while working at very slow rate, the UTM machine
vibration sometimes lead to very abrupt crack extension, which
might have also been due to improper notch tip preparation of the
specimens and hence the energy release rate value for those tests
are discarded. From the existing KIc values of literature paper of
[13], JIc values have been calculated using ASTM E1921 [48] while
 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225 209

Fig. 2. The SENB square cross-section (10
10 mm2) specimens having support length of 50 mm (a) SENB specimen with the fixture and (b) Enlarge view of notch tip as per ASTM
D7779 standard [49] (c) Schematic of fracture specimen.

approximating the values of Young's modulus of elasticity for the
different grades of graphite and tabulated in Table 2 (a). Interest-
ingly, from the calculation, significant differences between the GIc
and JIc values have been observed. Investigating the curiosity of
such mismatch, efforts have been made in the present work with
elaborate experimentation and computational simulations to find
or postulate some probable reason. As noted in Table 2 (b), the test
results, fracture experimentation and evaluation of GIc and JIc of
graphite specimens in this work also confirms significant difference
between these two parameters, though in all cases the limited
plasticity has been ignored and graphite is known to be a pre-
dominantly brittle material. Eventually, bimodularity has been
210 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225

Table 1 propagation region, thereby protecting the specimen. However,

Tension and compression test results for Young's Modulus of elasticity (ASTM E111). this results in a significant increase of test duration to one to 2 h for
Tensile Specimen ET (GPa) Compressive Specimen EC (GPa) a single specimen, which might be tolerated for bimodular brittle
Sample1 8.285 Sample1 5.582
specimens. This also results in a very smooth curve for load
Sample2 8.623 Sample2 5.653 displacement plot as the number of data points obtained are
Sample3 8.452 Sample3 5.624 significantly large. Fig. 3 demonstrates the modified load vs
Sample4 8.499 Sample4 5.591 displacement plots for all thirty successful SENB experimentation
Sample5 8.523 Sample5 5.673
for evaluation of fracture toughness and energy release rates.
Sample6 8.534 Sample6 5.431
Sample7 8.105 Sample7 5.745 Correspondingly, the peak load and fracture toughness for all
Sample8 8.384 Sample8 5.513 specimens have been determined and tabulated in Table 2 (a) and
Sample9 8.455 Sample9 5.534 (b). The strain energy release rate (G) has been calculated from
Sample10 8.683 Sample10 5.535
the experimental data point obtained from Instron UTM machine in
Sample11 8.577 Sample11 5.712
Sample12 8.469 Sample12 5.596
the form of load-displacement plot as per ASTM D7779-11 [49].
Sample13 8.578 Sample13 5.656 Following the initial nonlinearity at the start of the test, the load-
Sample14 8.607 Sample14 5.565 displacement diagram for notched bend specimen is character-
Sample15 8.484 Sample15 5.656 ized by a linear relationship. The data then deviate from a straight
Sample16 8.453 Sample16 5.672
line when crack propagation occurs. The first step in estimation of G
Sample17 8.589 Sample17 5.578
Sample18 8.747 Sample18 5.677 is a straight line plot for the initial portion of the load-displacement
Sample19 8.598 Sample19 5.689 plot. A linear straight line equation is formed to fit the data point
Sample20 8.399 Sample20 5.685 linearly as following equation
Average Value of ET 8.5022 Average Value of EC 5.61835

y ¼ mx þ b (2)

reasoned to be the candid factor for such difference in the fracture The data point is shifted in such a way that the straight line
parameter. The effect of bimodularity has been taken care of in the starts from the origin without changing the load value and the
fracture characterizing method via Energy release rate, because it is fitted line is expressed as:
based on load-displacement individual data point. The influence of
bimodularity has also been substantiated from the computational y ¼ mx (3)
results with unimodular and bimodular assumptions. Following The adjusted data point in load-displacement plot (P-D) due to
sections discusses the observations of the pertinent work. new shifted origin is considered in the crack-propagation. The
compliance for each data point, Cn is defined as
3.1. Bimodularity index from tension compression data set
Dn
The tensile Young's Modulus of elasticity ET ¼ 8.502 GPa and the Cn ¼ (4)
Pn
compressive Young's Modulus of elasticity EC ¼ 5.618 GPa are
evaluated according to ASTM E111 standard [52]. “Modulus of where, Cn ¼ Compliance for the point n (m/N).Dn ¼ Displacement
elasticity values for tension and compression are the corresponding for the point n (m), and.Pn ¼ Loading force for the point n (N).
average value estimated from the test results for all twenty speci- Then the crack lengthan , associated with each data point along
mens tested in tension and compression and the same has been the P-D diagram can be evaluated from the following expression.
tabulated in Table 1.” The graphite is found to be bimodular and its
bimodular index is determined as 1.51. an ¼ an1 þ ½ðW  an1 Þ=2*ððCn  Cn1 Þ=Cn1 Þ (5)

3.2. Evaluation of KIc, GIc and JIc where, a0 is the initial crack length of the SENB graphite specimen.
The strain energy release rate, Gðan Þ for each data point has been
The three point bend notch specimens are tested for evaluated from the following expression
S=Wratio ¼ 5 and a=W ¼ 0.5 (Fig. 2) within the specified range of h .  i
ASTM D-7779-11 [49]. The cross head speed for the progressive Gðan Þ ¼ P 2 2B *ððCn  Cn1 Þ=ðan  an1 ÞÞ (6)
fracture testing has been kept very slow at 0.005 mm/min, as the
material graphite is very brittle. Apart from obtaining large data Then, Gðan Þ is plotted againstDa, where Da ¼ an  a0 and shown
points for evaluating accurately the strain energy release rate G, the in Fig. 4. The critical strain energy release rate GIc is calculated from
slow cross head speed ensures reduced vibration in the crack G versus Da plot. The initial slope indicates the propagation of crack

Table 2
(a) Evaluation of JIc using published literature data [13] and comparison with GIc.

Specimen Type Fracture Parameter IG-110 IG-430 NBG-25-a NBG-25-c NBG-17-a NBG-17-c NBG-18-a NBG-18c PECA-a PECA-c

4.0T KIc (MPa.m1/2) 0.76 0.91 0.95 0.93 0.85 0.96 1.11 1.02 0.9 0.88
GIc (J/m2) 67 100 100 117 175 175 233 167 163 225
JIc (J/m2) 66.21 94.92 103.45 99.14 82.82 105.64 141.24 119.26 92.854 88.773

6.5T KIc (MPa.m1/2) 0.78 1.02 0.99 1.01 1.06 1.1 1.04 1.15 1.07
GIc (J/m2) 200 300 300 690 690 800 560 750 670
JIc (J/m2) 69.74 119.26 112.35 116.93 128.80 138.70 123.98 151.60 131.24

Here, 4.0T and 6.5T shows the two specimen size tested by Chi [13], where, 4.0T specimen means 50.0 (Length) x 10.0 (Depth) x 4.0 (Thickness) mm and 6.5T means specimen
52.0 (Length) x 12 (Depth) x 6.5 (Thickness) mm. ‘KIc’ is critical stress intensity factor, and ‘GIc’ critical strain energy release rate evaluated in (Chi 2016) [13]. ‘JIc’ is critical J-
integral value estimated from Eq. (7) with taking the assumed value of the Young's modulus of elasticity E ¼ 8.5 GPa and Poission's ratio is 0.16.
 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225 211

Table 2
(b) Evaluation of crack growth resistance in terms of fracture parameters KIc, JIc and GIc.

Sample No. Peak Load (N) Critical SIF (KIc) (MPa.m1/2) JIc (Critical J-integral value) (J/m2) Critical SERR (GIc) (J/m2)

Sample 1 81.81 1.097 137.91 169.87

Sample 2 90.56 1.2143 168.99 206.48
Sample 3 73.78 0.9893 112.17 132.72
Sample 4 79.4 1.0646 129.91 159.19
Sample 5 94.64 1.269 184.56 228.32
Sample 6 105.76 1.4181 230.48 289.88
Sample 7 83.95 1.1257 145.22 183.72
Sample 8 67.32 0.9027 93.385 125.9
Sample 9 91.89 1.2321 173.99 219.53
Sample 10 67.15 0.9004 92.914 112.43
Sample 11 103.78 1.3916 221.93 267.54
Sample 12 100.51 1.3477 208.16 257.88
Sample 13 68.3 0.9158 96.124 116.31
Sample 14 99.67 1.3364 204.7 249.63
Sample 15 86.67 1.1621 154.78 189.21
Sample 16 82.88 1.1113 141.54 176.29
Sample 17 101.78 1.3647 213.46 256.28
Sample 18 98.63 1.3225 200.45 249.94
Sample 19 97.78 1.3111 197.01 238.39
Sample 20 85.71 1.1493 151.37 189.16
Sample 21 94.07 1.2614 182.34 223.61
Sample 22 97.21 1.3035 194.72 236.41
Sample 23 88.23 1.183 160.41 199.02
Sample 24 96.54 1.2945 192.05 235.23
Sample 25 98.89 1.326 201.51 249.82
Sample 26 88.56 1.1875 161.61 199.57
Sample 27 94.78 1.2709 185.11 227.94
Sample 28 91.89 1.2321 173.99 214.57
Sample 29 91.78 1.2306 173.57 211.02
Sample 30 93.33 1.2514 179.49 219.67

Fig. 4. The plot between energy release rate and crack extension.Da
Fig. 3. Load versus displacement plot for all thirty SENB specimens.

corresponding JIc values calculated from the fracture toughness

and the strain energy release rate may reach a plateau indicating
steady state crack propagation, or tend to decrease with further
crack extension. If there is upward inflection in the G versus Daplot
as the specimen nears failure with a maximum occurring at failure,
the initial peak in G is reported as the critical strain energy release
rate, GIc , for that particular test result.
Fig. 4 delineates the variation of strain energy release rate G
versus incremental crack growth Da for all the thirty successful
experimentation. The retrieved energy release rate plot patterns
are though similar, but has large scale scatter in GIc values.
The test specimens having distinct value of GIc has been calcu-
lated as the maximum value of first peak observed in the plots and
tabulated in Table 2(b). The GIc values are compared with the
employing the following expression taken from ASTM E1921-17a
[48]:
  2
1  n2 KIc
JIc ¼ (7)
E
where, JIc is the critical J-integral value which is equivalent to
critical energy release rate value for brittle material with no
permissible growth of plastic zone. KIc is the critical stress intensity
factor calculated from ASTM D-7779 with experiment [49]. E is
Young's modulus of elasticity in tension. The JIc value is calculated
for comparing the unimodular and bimodular behaviour which are
also appended from the numerical and stochastic simulations. From
212 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225

Table 2 (a) and (b), significant difference between critical J-integral  

value JIc and critical strain energy release rate GIc has been Pmax m
Pi ¼ 1  exp  Pmax > 0 (9)
observed. This leads to performing further computational analysis sq
for establishing the influence of bimodularity on the fracture
behaviour of graphite components. This can be reasoned here that Pi ¼ 0Pmax  0 (10)
the critical stress intensity factor KIc has been calculated using the
peak load which does not include the inherent effect of bimodu- where, Pi is probability of failure, Pmax is Peak load attained during
larity at each data points. However, the critical strain energy release experiment, sq is Weibull scale parameter or characteristic strength
rate evaluation requires G versus Da plot for the individual data at 63.21% of probability of failure and m is slope parameter or
points, which has been evaluated from the SENB experimentation Weibull modulus. The derivative of Eqs (9) and (10) gives the
using digital image correlation set up with the UTM machine. Then expression for the probability density function and is written as:
pixel by pixel image processing of crack growth data retrieving is
   

possible for reliable and accurate plotting of energy release rate m Pmax m1 Pmax m
f ðPmax Þ ¼ exp  Pmax > 0 (11)
plots for the SENB specimens. sq sq sq

3.3. Weibull model of fracture experimentation f ðPmax Þ ¼ 0Pmax  0 (12)

The fracture testing of thirty single edge bend notched speci- where, f ðPmax Þ is probability density function. Fig. 6 displays the
mens gives scattered data set for the peak load obtained from load- probability density function f ðPmax Þ plotted against the Peak load.
displacement curve. Stochastic analysis of these randomly distrib- Weibull theory correlates the strength distribution parameters
uted data has been carried out to characterize graphite component based on specimen geometry to a strength distribution parameter
failure. For a reliable design, exact characterization of the crack based on the material property. As the size of a component or test
growth parameter is necessary for the life prediction of the graphite specimen geometry is increased, then, on an average, the tensile
component part. In the case of cracked specimens, there is no need strength of the component decreases. The reason for this is that as
for the final fracture of the material as the fracture toughness is the volume (or surface area) of the component is increased, the
characterized by the peak load. Fracture toughness and critical likelihood of encountering a critical flaw with deleterious orien-
strain energy release rate values are compared with numerical tations to the load applied increases. The estimated distribution
simulation results to analyze the effect of bimodularity. The prob- parameters are obtained by taking the logarithm of Eq. (9) and after
ability of failure of component has been evaluated as a function of rearranging the expression becomes
fracture strength using Weibull statistical model. The random peak
load data is plotted with help of two and three parameter Weibull lnðlnð1=ð1  Pi ÞÞÞ ¼ m lnðPmax Þ  m lnðsq Þ (13)
distribution using least square and maximum likelihood estima- Eq. (13) is in the form of equation of a straight line, i.e.,
tors. The success of fitting the Weibull distribution through the
normalized data sets allows us to improve the basis for the esti- Yi ¼ aXi þ c (14)
mates of the variability. This could also imply that the variability on
the graphite strength for the different strength measures is based where, Yi ¼ lnðlnð1=ð1  Pi ÞÞÞ,
on the same flaw distribution and thus a property of the material.
Since fracture toughness is considered a material property as the a ¼ m ðslopeÞ;
resistance against progressive crack propagation, therefore a
unique fracture toughness value has to be predicted for the mate- Xi  lnðPmax Þ;
rial. This has been determined from the probability of failure versus
peak load distribution plot of the Weibull model of experimental c ¼ m lnðsq Þ
data sets as shown in Fig. 5. To estimate the unique fracture
toughness for the material, Weibull distribution plot for cumulative Using the traditional linear regression, a and c are obtained from
probability of failure is plotted against the peak load. The cumu- the following expressions:
lative probability of failure is obtained by the following ranking
scheme as per ASTM C 1239 [53].
P
n P
n P
n
n Xi Yi  Xi Yi
i¼1 i¼1 i¼1
i  0:5 a¼ !2 (15)
Pi ¼ (8) P
n
2 P
n
n n ðXi Þ  Xi
i¼1 i¼1
where Pi is the ranked cumulative probability, i is the rank of the
individual specimen arranged in increasing order of peak load and P
n P
n P
n P
n
n is the total number of specimens tested. The scattered data on the ðXi Þ2 Yii  Xi Xi Yi
Weibull plot has been analyzed using the best fit line drawn by least c ¼ i¼1 i¼1 i¼1 i¼1
!2 (16)
square estimator (LIN2), biased (MLE2-B) and unbiased (MLE2-U) P
n
2 P
n
maximum likelihood estimator procedure. n ðXi Þ  Xi
i¼1 i¼1
Linear regression using least square estimator requires the
application of a ranking rule for all the failure load data to estimate Once ‘a’ and ‘c’ are estimated from regression, the Weibull pa-
the failure probability of each data. Once an estimate of individual rameters m and sq can be extracted from the expressions in Eq. (13).
probabilities is made, the failure data is regressed to determine the Graphically the value of m is the slope of Eq. (13), and sq is the value
distribution parameters. Finding distribution parameters via linear of peak load corresponding to the 63.2% probability of failure (refer
regression involves utilizing the ranked probability of failure along to Fig. 5).
with the associated failure stresses in the following expression The Maximum likelihood estimators are superior to linear
based on Weibull weakest link theory [54,55]: regression methods for determining the value of m from a set of
 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225 213

Fig. 5. Weibull distribution plot between Cumulative Probability of failure of cracked specimens against peak load and best fit line drawn with least square estimator (LIN2), biased
(MLE2-B) and unbiased (MLE2-U) maximum likelihood estimator.

fracture strengths. The maximum likelihood estimators (MLE) yield

unique solutions for the distribution parameters for a two- P
n
ðPmaxi Þm lnðPmaxi Þ
parameter Weibull distribution. As the sample size increases,
i¼1 1Xn
1
these estimators asymptotically converge to the true distribution  lnðPmaxi Þ  ¼ 0 (18)
P
n n m
parameters and the confidence interval narrows as well relative to ðPmaxi Þm i¼1
i¼1
other estimators. This latter fact makes this a better choice of the
estimation scheme. Let Pmax1, Pmax2, …., Pmax n represent Peak load
for different experiments (a random variable) in a sample data set, ! !1=m
X
n
1
where it is assumed that the Peak Load is characterized by the two- sq ¼ ðPmaxi Þm (19)
parameter Weibull distribution. The likelihood function associated i¼1
n
with this sample is the cumulative probability density evaluated at
Eq. (18) is solved in iterative fashion, for estimation of m. Sub-
each of the ‘n’ sample values. This likelihood function is dependent
sequently, sq is estimated from Eq. (19) using the value of m
on the two unknown Weibull distribution parameters (m, sq ). The
calculated from Eq. (18). A closed form solution for Eq. (17) is not
likelihood function, L, for an uncensored sample under these as-
available, thus this expression must be solved numerically and has
sumptions is given by the expression:
been plotted in Fig. 5 using WeibPar 4.3 software (Procured from
Connecticut Reserve Technology). The material is considered safe
before attaining the peak load and above this value considered to
be failed. The probability of failure against peak load data set has
been plotted with 90% confidence bound as shown in Fig. 5. The
probability plot shows a graph with observed cumulative per-
X n  
X  

n
m Pmaxi m1 Pmaxi m centage on horizontal axis and expected cumulative percentage on
L¼ f ðPmaxi Þ ¼ exp 
i¼1 i¼1
sq sq sq vertical axis. The Weibull probability plot is used to test whether or
not a dataset follows Weibull distribution. The abscissa and ordi-
(17)
nate of the distribution plot are scales of type Loge and Double Loge
For an uncensored sample data the system of equations is ob- Reciprocal respectively. If all the scatter points are close to the
tained by differentiating the log likelihood function with respect to reference line, we can say that the dataset follows the Weibull
m andsq . The derivatives of Eq. (17) is equated to zero and repre- distribution. The horizontal axis is expanded within the range limit
sented as: of peak load found from experiments for better interpretation of
214 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225

Fig. 6. The probability density function plot with random data representation in bar chart with normal and Weibull distribution.

the plot. Maximum Likelihood Estimator Unbiased (MLE2U) has been used
The value of co-efficient of determination is found to be more in the subsequent section for evaluating characterizing fracture
than 0.95 which may improve by increasing the number of speci- parameters KIc and JIc values.
mens tested. The variation of mean peak load determined from all
distribution estimators are within a small range. The value of mean
peak load evaluated from the unbiased maximum likelihood esti- 3.4. Formulation of path independent J-Integral
mator for the Weibull two parameter distribution has been further
used to find the critical stress intensity factor of the graphite The investigation of bimodular fracture problem qualify to be a
component. Fig. 6 represents the histogram plot between the fre- stress dependent elasticity problem necessitating a rigorous step-
quency (probability density function) and peak load. This plot ped iterative evaluation of state of stress for each iteration to
shows the Weibull distribution with least square estimator and determine the tensile and compressive stresses. Three-dimensional
maximum likelihood estimators and compared with normal dis- computational methods based on finite element procedures have
tribution. The random peak load data is plotted in the square bar been adopted to evaluate the path independent fracture parameter
chart form and it is clearly visible that normal distribution cannot denoted as 3D J-integral. This requires the determination of line and
fit the failure data set. The Weibull parameters and goodness of fit area integrals iteratively with programming schemes which assign
(R2) for Weibull distribution estimated through the least square a definitive value to modulus property based on positive or nega-
estimator and maximum likelihood estimator have been tabulated tive state of stress, which eventually becomes the starting point for
in Table 3. next iteration till desired convergence and computational accuracy
The mean peak load 94.282 N evaluated from the two parameter is satisfied. Though both domain integral and contour integral
methods can be effectively used for deriving 3D J-integral for
unimodular and bimodular material system, however in the pre-
Table 3 sent work, the contour integral method has been followed for its
Co-efficient of determination (R2), slope parameter and mean characteristic load ease and simplicity of incorporating into numerical programs and
parameter with three parameter and two parameter Weibull distribution.
macro subroutines for evaluating the J-integral. For completeness
R2 Value slope m mean peak load Threshold load of the study, expressions for unimodular path independent integral
LIN2 0.96011 10.2329 94.351 has also been provided. The state of the work presented here gives a
MLE2B 0.95159 11.2205 94.273 detailed analytical and numerical correlation of path independent
MLE2U 0.95815 10.6935 94.282 integral for the stress dependent elasticity problem of fracture in
LIN3 0.96011 10.2329 94.351 3.42E-06 bimodulus material media. In the absence of closed form solution
MLE3 0.95159 11.2205 94.273 4.55E-06
for such category of problem due to its inherent complexity of
 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225 215

modulus being a function of state of stress, finite element proced-

ures have been adopted for describing the fracture parameter of  
cracked numerical specimens. ∭ sij u_ i ;j dV ¼ ∬ sij u_ i nj dS ¼ ∬ Ti u_ i dS (24)
V S S

3.4.1. Bimodular formulation of path independent 3D J-Integral Introducing Eq. (24) into Eq. (22) and using strain derivatives
Fig. 7 shows the schematic of a plate containing a crack for the _ ij for displacement derivative, it is shown that
2
purpose of quantifying path independency in 3D domain. Now, the
crack tip is assumed to virtually move an infinitesimally small
 
distance from the fixed frame at O to moving frame at O1. The di- ∭ sij;j u_ i dV ¼ ∬ Ti u_ i dS  ∭ sij ε_ ij dV (25)
rection of X2 and x2 are perpendicular to the crack surface corre- V S V
sponding to O and O1 respectively. Two contour paths are chosen, Now, Eq. (21) can be written as
the first one being the outer contour noted by path G1 curves and
the second contour G2 can be any arbitrary contour surrounding the
crack surfaces. The region enclosed by these two contours is A1 and ∬ Ti u_ i dS þ ∭ Fi u_ i dV ¼ ∭ ru€i ui dV þ ∭ sij ε_ ij dV (26)
the area bounded by the crack plane and the second contour isA2 . S V V V
Now stating the equilibrium equations for a stressed continuum
For an infinitesimal virtual crack extension, the energy release
of volume V subjected to arbitrary traction Tand body forces F
rate can be evaluated from Eq. (26) for a differential change dl of the
sij;j þ Fi ¼ ru€i (20) propagating crack as similar to time derivatives. This is expressed as

where, sij , Fi , r and ui are stress-tensor, body force per unit volume, Z
density of solid and displacement respectively where the first and dui du du dεij
Ti dG þ ∬ Fi i dA ¼ ∬ ru€i i dA þ ∬ sij dA þ J u
second dots specify the first or second time derivative of dl A1 dl A1 dl A1 dl
G1 þGc
displacement. Multiplying u_ i on both sides of Eq. (20) and inte-
grating over the body volume V (27)
 
∭ sij;j þ Fi u_ i dV ¼ ∭ ru€i u_ i dV (21) where J u is the rate of change of energy of material in the fracture
V V process region for name say a generalized universal integral, be it
unimodular or bimodular. We can introduce zero integral terms
Now expanding ðsij u_ i Þ;j ¼ sij;j u_ i þ sij u_ i;j , the first part of the in-
with reference to contour G2 with an integral evaluated along the
tegral of Eq. (21) after rearrangement
contour path and opposite the contour path by adding and sub-
  R
∭ sij;j u_ i dV ¼ ∭ sij u_ i ;j dV  ∭ sij u_ i;j dV (22) tracting the term Ti dudl
i
dG to Eq. (8) as
G2
V V V

Now traction on a small differential element on the contour Z Z

dui dui du
surface can be expressed as Ti dG þ Ti dG þ ∬ Fi i dA
dl dl A1 dl
G2 G2
Ti ¼ sij nj (23) Z
du dεij dui
¼ ∬ ru€i i dA þ ∬ sij dA þ J u  Ti dG (28)
where, nj is the outward normal vector. A1 dl A1 dl dl
G1 þGc
Substituting traction form into the integral equations formu-
lated above and using Gauss's Theorem (divergence theorem), it Upon rearrangement following expressions for energy release
can be written as rate J u is obtained.

Fig. 7. Configuration of crack tip {G1 (arbitrary curve surrounding area A1 ), Gc (curve along the crack surface), A2 (fracture process region), G2 (boundary of A2 )} around a region of
infinitesimal thickness enclosing the Crack Front.
216 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225

Z Z   
dui dui du dεij
Ju ¼ Ti dG þ Ti þ ∬ ðFi  ru€i Þ i  sij
dl dl A1 dl dl
G1 þGc G2 G2

dA

(29)

  Z   
du du du dεij
J u ¼ ∬ sij i dA þ Ti i dG þ ∬ ðFi  ru€i Þ i  sij
A1 dl ;j dl A1 dl dl
G2

dA

(30)
2
6
Ju ¼ ∬ 4 (31)
A1

Using Eq. (20), the first term of area integral in Eq. (31) vanishes
and the integral may be written as
Z
dui
Ju ¼ Ti dG (32)
dl
G2

With reference to Fig. 8, transformation equations from the

fixed frameO  X1 ; X2 to moving frame O1  x1 ; x2 for the infini-
tesimal crack extension can be expressed as

x1 ¼ X1 cos q0 þ X2 sin q0  l
(33)
x2 ¼ X1 sin q0 þ X2 cos q0

X1 ¼ x1 cos q0  x2 cos q0 þ l cos q0

(34)
X2 ¼ x1 sin q0 þ x2 cos q0 þ l sin q0
Similarly, displacements for the fixed frame O  X1 ; X2 is given Fig. 8. Representation of propagation of crack tip from O to O1 in (a) 2D boundary and
by (b) 3D domain.

ui ðX1 ; X2 ; lÞ ¼ ui ðx1 cos q0 þ x2 sin q0  l; x1 sin q0 dui vui vui vxi vui vui vx1 vui vx2
¼ þ ¼ þ þ (36)
dl vl vxi vl vl vx1 dl vx2 vl
þ x2 cos q0 ; lÞ (35)

Now substitution of Eq. (38) into Eq. (32) gives, vui vx1 vu vx
¼  i; 2 ¼ 0 (37)
Z   vx1 vl vx1 vl
vui vui
Ju ¼ Ti  dG (39)
vl vx1
dui vui vui
G2 ¼  (38)
dl vl vx1
Here, J u -integral is the crack driving force or also known as the
energy release rate during crack extension.
We assume that the fracture process region does not depend
upon load conditions or upon geometry of body or crack. Hence, the
process region is assumed to be constant in dimensions and moving
along with the same speed as the crack tip, and hence,vu vl
i
¼ 0 holds Z  
vu vu
inG2 . Ju ¼  Ti cos q0 i þ sin q0 i dG (42)
Then, Eq. (39) is simplify to vX1 vX2
G2
Z  
vui For any arbitrary orientation q0 of the propagating crack front,
Ju ¼  Ti dG (40)
vx1 the J u can be resolved as
G2

We obtain from Eqs. (14)e(16)

J u ¼ J1u cos q0 þ J2u sin q0 (43)
vui vu vu
¼ cos q0 i þ sin q0 i (41) Taking as single notation Jku
where k ¼ 1, 2 correspond to
vx1 vX1 vX2
respective coordinate axes, Eq. (40) can be modified using Eq. (43)
Substituting Eq. (41) in Eq. (40) the integral equation becomes as
 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225 217

Z However, when area of the fracture region A2 is very small, the

vui
Jku ¼  Ti dG (44) contour integral G2 become approximately equal to zero and then
vXk Eq. (55) can be written as
G2
! Z
However, for verification of that the above integral is path in- vεuij
dependent, let us consider another integral surrounding the crack ∬ sij dA ¼ W u nk dG (56)
A1 vXk
path and expressed as G1 þGc
Z
u vui For a general class of brittle solid, the two dimensional form of
Jk ¼  Ti dG þ Mk ðAÞ (45) contour integral can now be expressed as
vXk
G1 þGc
Z  
vui vui vεij
where, Mk ðAÞ are the terms determined when the area A1 sur- Jku ¼ ∬ ðru€i  Fi Þ dA  Ti dG þ ∬ sij dA (57)
A1 vXk vXk A1 vXk
rounded by G1 ; G2 and Gc is specified. Now, if both the integral Jku and GþGc
u
J k are different, then we can write
Substituting Eq. (56) into Eq. (57), we obtain
Z Z
u vu vu Z 
Jk  Jku ¼  Ti i dG þ Mk ðAÞ þ Ti i dG (46) vu vu
vXk vXk Jku ¼ W u nk  Ti i dG þ ∬ ðru€i  Fi Þ i dA ðk ¼ 1; 2Þ
G1 þGc G2 vXk A1 vXk
G1 þGc
Z (58)
u vui
J k  Jku ¼  Ti dG þ Mk ðAÞ (47)
vXk Let us express the above Eq. (58) as
G1 þGc G2
 
Z Jku ¼ Jku 2D (59)
u vu
Jk  Jku ¼  Ti i dG þ Mk ðAÞ (48)
vXk Now with reference to Fig. 7, the three dimensional form of path
Gt independent J-integral can be defined by considering the two faces
of the area (A ¼ A1 þ A2 ) as Aþ andA and then integrating the 2D J
Gt ¼ G1 þ Gc  G2 denotes the contour which surrounds the area A1. integral through the elemental thickness dh. Since the two face-
u
For path independence J k ¼ Jku and hence from Eq. (48)
sAþ andA are parallel and in opposite directions, the sum of the
Z   two area integral is given by the X3derivative of the integrand
vui vu
Mk ðAÞ ¼ Ti dG ¼ ∬ sij i dA (49) multiplied bydh. Thus, the 3D J-integral is the addition of contour
vXk A vXk ;j
Gt integral evaluated over a remote contour and the area integral
evaluated over the two faces enclosed by the contour. This is given
Therefore, as
  Z
u vu vui  u   v  u
J k ¼ Jku ¼ ∬ sij i dA  Ti dG (50) Jk 3D ¼ Jku 2D þ ∬ Jk 2D :n3 dA (60)
A1 vXk ;j vXk vX
G1 þGc A 3

Using Eq. (20), we can write Eq. (50) as Now, 2nd term of right side of above equation is expressed as
   9
u vu vεij  = 
J k ¼ ∬ ðru€i  Fi Þ i þ sij dA v  u v vu v
vXk vXk ∬ Jk 2D :n3 dA ¼ ∬ W u nk  Ti i n3 dAt þ ∬ ru€i
A1
Z A vX3 vX3 A vXk ; vX3 A
vu
 Ti i dG ðk ¼ 1; 2Þ (51)  vu
vXk  Fi i
:n3 dAt
GþGs vXk
Let us consider a purely brittle elastic material with negligible
(61)
plastic zone during crack front propagation. The strain tensor εij for
such material is denoted as All the term except 1st term of Eq. (61) will vanish as variation of
other terms in the direction n3 is constant. Then it can be given as
εij ¼ εuij (52)
 
v  u v v vu
εuij
is the elastic strain field and Wu
∬ Jk 2D n3 dA ¼ ∬ W u nk n3 dA  ∬ sij nj i n3 dA
A vX3 A vX3 A vX3 vXk
where, is the elastic strain en-
ergy density function and hence
(62)
v
W u ¼ sij (53) Using Eq. (62) in Eq. (60), we obtain
vεuij
Z  
!  u vui v
  vεuij Jk 3D ¼ W u nk  Ti dG þ ∬ W b nk n3 dA
vεij vXk vX
∬ sij dA ¼ ∬ sij dA (54) G1 þG2 þGc
A 3
A1 vXk A1 vXk  
v vu vu
! ∬ sij nj i n3 dA þ ∬ ðru€i  Fi Þ i dA (63)
Z vX 3 vX vX
vεuij A k A k
∬ sij dA ¼ W u nk dG (55)
A1 vXk For any arbitrary crack front progression [Fig. 8(b)] behaviour
G1 þGc G2 the 3D path independent integral is expressed as
218 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225

 u       sT sC
Jk 3D ¼ J1u 3D cos q0 þ J2u 3D cos F þ J3u 3D cos J (64) εT ¼ and εC ¼ (68)
ET EC
where, q0, F and J is the orientation of front with respect to X1, X2
where, subscripts ‘T’ and ‘C’ denote the constituent tensile and
and X3 axes respectively. For example, crack propagation variation
compressive parameters and s, ε, and E respectively represent the
dominant in one direction (k ¼ 1) and very small in other two di-
stress, strain and Young's modulus for a given material and loading
rections (k ¼ 2,3), the J-integral can be expressed as
configuration. Since the modulus of elasticity itself is a function of
 u   state of stress and the latter being an unknown as a priori a step
Jk 3D y J1u 3D cos q0 (65) function approach is adopted as similar to a Dirac delta function
having a value of either one or zero depending on the state of stress.
Now from Eq. (63), this can be evaluated as
The model step function of the bilinear bimodular material for
8 9 representing the stress dependent elasticity problem can be writ-
Z < = ten as
 u vu v
J1 3D ¼ W u n1  Ti i dG þ ∬ W u n1 n3 dA
: vX1 ; vX 00  1 0  11
A 3
G1 þG2 þGc
  @ @U sb A @U  sb AA b
v vu vu b
ε ¼ þ s (69)
∬ sij nj i n3 dA þ ∬ ðru€i  Fi Þ i dA ET EC
A vX 3 vX1 A vX1

(66) where, U is a step function defined as

Further simplifying Eq. (66), the 3D path independent integral is

expressed as
UðεÞ ¼ 0 if ε < 0;
(70)
¼ if ε > 0
Z  The superscript b implies bimodularity andεb and sb represent
 u vu  
J1 3D ¼ W u n1  Ti i dG  ∬ si;3 ui;3 ;3 dA þ ∬ ðru€i the bimodular strain and stress field respectively. The constitutive
vX1 A A
G1 þG2 þGc model for the three-dimensional tensile state of stress is given as
vui
 Fi Þ dA
vX1 εTij ¼ aTijkl sTkl (71)
(67)
and for compressive region
When the material exhibits bimodularity, the stress dependent
elasticity formulation has to be incorporated into Eq. (67) for
evaluating J-integral. As the material experiences different stress εCij ¼ aCijkl sCkl (72)
strain behaviour under tension and compression as shown in Fig. 9,
Bimodular constitutive relationship is given by
the stress field defining stress dependent elasticity phenomena has
to be modified accordingly. The linearized and non-linearized
constitutive model exhibiting different elastic moduli in tension εbij ¼ abijkl sbkl (73)
and compression requires iterative numerical procedures for eval-
uating the state of stress in the cracked specimen. The present work The flexibility coefficient tensor,abijkl for bimodular state of stress
follows the bimodular Ambartsumyan constitutive model [1e7] for is defined as
quantifying the crack tip stress field. The bimodular theory postu- h  i h  i
lates the tension-compression stress strain plots into two straight abijkl ¼ aTijklmn U fmn sbpq þ aCijklmn U  fmn sbpq (74)
lines, whose tangents at the origin are discontinuous. This seems
adequate for efficient modeling of the fracture behaviour of many εTij , εCij and sTkl , sCkl are strain tensor and stress tensor in tensile and
natural and man-made brittle materials demonstrating bimodu- compressive region respectively. aTijkl , aCijkl are the respective flexi-
larity. The elastic continuum mechanics approach has the bility coefficient tensor in tensile and compressive state. The
assumption that the stressed body is continuous, homogeneous, functions fmn are the functions of state of stress. The constitutive
and isotropic and all deformations are small enough to have theory equations so developed has to satisfy the principle of coordinate
of superposition valid for all response characteristics. invariance. Therefore, fmn are independent of coordinate rotation
The bimodular constitutive model for tensile and compressive being isotropic functions of their arguments. Hence, the function
state of stress has been defined as fmn is expressed as an argument as follows.

Fig. 9. Bimodular constitutive model: (a) Bilinear model when ET > EC (b) bilinear model when ET < EC and (c) non-linear model for actual condition.
 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225
fmn ¼ fmn ðI1 ; I2 ; I3 Þ
where, I1 ,I2 and I3 are the well-known stress invariant of elasticity
theory, which satisfy the following cubic equation.
s3  I1 s2 þ I2 s2  I3 ¼ 0
fmn ¼ 0 if msn and
fmn ¼ si if m ¼ n
fmn ðspq Þ ¼ sr where, r ¼ I, II, III are the three roots of Eq. (76). Since
the bimodular elastic constants are the functions of the sign of
principal stresses, this can be incorporated to Eq. (73) for defining
the general form of bimodular elasticity. So we obtain
h i
εbij ¼ aTijklr Uðsr Þ þ aCijklr Uð  sr Þ sbkl
where, the εbij , sbkl are the bimodular tensorial strain and stress
matrix, whereas sr is function of principle stress. However, with
reference to the principle of coordinate invariance for any arbitrary
plane, the mean principal stress or octahedral normal stress so is
considered as a function of sr and the step function with reference
to Eq. (70) is defined as follows.
219
redundant. The rigors of bimodular crack tip stress field formula-
(75) tion to characterize the three dimensional crack growth behaviour
might preclude its adoption to elementary structures, but consid-
ering the unprecedented failure uncertainties in high risk struc-
tures, the catastrophe can be better designed with a certain degree
of reliability gaining over all the computational effort. The
(76) following section elaborates the numerical examples for evaluation
of path independent integral in a bimodulus brittle material.
For evaluation of 3D path independent integral J, single edge
cracked bend specimens of supposed bimodular brittle material
77 with negligible crack tip plastic zone has been considered. The
three-point bend specimen with centrally located through thick-
ness edge crack of radius ‘a’ with geometry, loading and boundary
condition as shown in Fig. 10 has been used in the numerical
simulation for the bimodular elasticity crack field problem. It has
been presumed that depending on the state of stress the elastic
property of the specimen changes based on the tensile and
compressive nature of the state of stress. Altogether 12 different
(78)
contours around the crack front for each cracked bend specimen
have been chosen for evaluating fracture parameter J. The signifi-
cant number of numerical experimentation is justified as the
fracture parameter is very sensitive to bimodular index, which
might have been avoided for non-fractured studies with specimens
devoid of any crack like defect.
A detailed parametric study has been carried out for assessing
the influence of modulus ratio upon fracture characteristics of
  bimodular flexure specimen by evaluating surface and area in-
s þ s2 þ s3
UðxÞ ¼ 0 if x ¼ 1 < 0; tegrals over ten contours surrounding the crack path. Such short
3 (79) comings of 2D J-integral with an unrealistic plane stress and plane
¼ 1 if x > 0 strain assumption for an inherently three dimensional bimodular
problem has been effectively tackled by considering the parallele-
The strain tensor so developed being the function of octahedral
piped elements across the crack front. Thereby the error due to
normal stress field is expressed as
dependence of crack to width ratio and thickness effect for an
h i asymmetric zero stress state surface has been neutralized. The
εbij ¼ aTijklo Uðso Þ þ aCijklo Uð  so Þ sbkl (80) three dimensional bimodular J-integral formulation as derived in
Eq. (84) has been computed for each bimodular index for all ten
1 contours for a single cracked specimen. The constitutive integral of
W b ¼ sbij εbij (81) overall 3D J-integral derived is abbreviated in the following form:
2
 
1 h i
J1b ¼ JLb  JAb (85)
W b ¼ sbij aTijklo Uðso Þ þ aCijklo Uð  so Þ sbkl (82) 3D
2

where, so is the octahedral normal stress and W b is the strain en- where, JLb is the line integral equivalent to Rice's Integral, evaluated
ergy density of bimodular elasticity. The traction, Tib is expressed over a contour that lies in the principal normal plane of the crack
for bimodular stress field as front that encloses the crack tip and JAb denotes an integral evaluated
over the area enclosed by the contour. Appropriately, program
Tib ¼ sbij nj (83) subroutines have been written to incorporate the above condition
in the finite element programming software COMSOL to model the
Eq. (67) has been modified by using Eqs. (78), (80), (82) and (83) stress dependent elasticity problem of crack progression behaviour
to derive the path independent J-integral for the bimodular stress iteratively. The detail of the programing flowchart has been shown
field to characterize crack progression behaviour. Thus we get, in Fig. 11. The computational time on the server based parallel
 computing depends upon the tolerance accuracy of stress depen-
  Z
1 b T  dent elasticity.
J1b ¼ s a Uðso Þ þ aCijklo Uð  so Þ sbkl n1
3D 2 ij ijklo The first iteration has been solved assuming the material to be
G1 þG2þGc isotropic and unimodular. The neutral surface and geometric
  
vui
 sbij nj dG  ∬ sbi3 ui;3 dA þ ∬ ðru€i  Fi Þ i dA
vX1 A ;3 A
Now the exact expression for J-integral becomes complex
requiring efficient computational tools to evaluate the fracture
parameter. The phenomenon of bimodularity is predominantly
observed in brittle materials, which has negligible growth of crack
tip surrounding plastic zone during crack propagation. Therefore,
using elastic-plastic analysis for bimodular stress field seems
vu symmetric mid-surface both located at the same place. Then the
vX 1 iteration invokes the step function to define the bimodularity based
on the sign of function ‘so ’ and then reevaluation of the state of
(84)
stress is carried out. The do loop is carried forward till the
convergence criteria is satisfied. The accuracy of this three-
dimensional problem for determining neutral surface iteratively
is very much mesh sensitive and depends on the number of sub
steps for each iteration. Fig. 11 demonstrates this logic flow chart for
the bimodular flexural problem. The flow chart starts with location
of neutral axis for taking the tensile modulus as unimodular stress
220 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225

Fig. 10. Contour representation for evaluating J-integral.

field. According to the sign of the principal stress function param-
eter, the domain of the tensile stress filed has been identified
separately from the compressive stress field for each subsequent
iteration. Subsequently, the tensile and compressive modulus of
elasticity is assigned to the tensile and compressive region. The
bimodular effect now can be observed in the flexural simulation of
the cracked specimen due to non-uniform tensile and compressive
zones. In the next iteration the region of tensile and compressive
zone again changes and so also the neutral axis is shifted. The
procedure is repeated till a predefined tolerance limit of close to
0.001 has been achieved. The change of state of stress between two
consecutive iterations if remain constant for 99.99% of the assigned
nodes, then the iteration is stopped. To lessen the computational
time, advantage of parallel processing has been invoked for each

Fig. 11. Program flow chart for evaluation of neutral surface in a bimodular flexure specimen.
 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225
bimodular index.
3.5. Computational analysis
Finite element (FE) simulations have been performed for the
unimodular (ET/EC ratio ¼ 1) and bimodular (ET/EC ratio ¼ 1.51 ob-
tained from experiment) numerical three point bend smooth edge
cracked specimens using mean peak load (94.282 N) as the loading
condition. The generalized loading, crack geometry, analytic and
finite element model for a widespread ET/EC ratio has been dis-
cussed elsewhere to study the effect of bimodularity on the fracture
parameter J-integral. Finite element model of the numerical spec-
imen has been created as similar to the experimental specimen
used in the SENB fracture testing. With reference to earlier Fig. 2
(c); the S/W ratio is 5 and a/W is taken to be 0.5 as similar to
ASTM D7779 standard [49]. The FE meshing for the similar geom-
etry between the support to the actual experimental test geometry
and subjected to the mean peak load obtained from Weibull two
parameter unbiased maximum likelihood estimator is shown in
Fig. 12. The model has been used for comparing the critical stress
intensity factor with unimodular and bimodular consideration of
the material. Scientifically judicious mesh grading and refinement
has been carried out along the crack tip region to capture the crack
tip stress field. Appropriate convergence studies have been carried
out to optimize the computational effort and time without
compromising on the analysis results. Typically the meshing shown
in Fig. 12 comprises of 20990 hexahedral elements. The effect of
bimodularity is prominent in y-direction, so, in this direction 20
divisions has been taken for better estimation, whereas, other di-
rection (z-direction) only 10 divisions are taken, because it is in-
dependent of bimodularity.
The plot of Von-Mises stress shown in Fig. 13 which clearly
shows the stress is very high in the crack-tip region. Another plot
Fig. 12. FE mesh distribution for the edge cracked three-point bend equivalent numerical specimen.
221
shown in Fig. 14 for Young's Modulus, distinguishes the tensile and
compressive region over the whole domain. On the basis of the
input ingredients, the simulation of the 3D J-integral were per-
formed and the value of J-integral and constitutive integral (Line
integral or Rice's J-integral and Area integral for 3D geometry ob-
tained from derivation) for twelve contours. There are very
distinguish effect of bimodularity has been seen while, comparing
the unimodular and bimodular results. The interesting point is that
the unimodular results are approaching toward the KIc value
calculated from Eq. (1) with taking the value of peak load whereas
the bimodular values are approaching toward the G value calcu-
lated from Eq. (2) with taking the load vs displacement instanta-
neous data points.
Fig. 15 (a) and (b) shows that the comparison of 3D J-integral
value with its constituent line and area integral for bimodular and
unimodular stress field respectively. In Fig. 16, two curves are
plotted for GIc and JIc against the peak load obtained in load-
displacement plot. Then corresponding to the mean peak load
estimated from the Weibull distribution, the critical fracture pa-
rameters JIc and GIc have been evaluated for the bimodular graphite
specimen. The specific crack parameter is evaluated at the partic-
ular mean peak load from the above curves by drawing a vertical
straight line at the mean peak load and the corresponding inter-
section points give the specific JIc and GIc Value which has also been
tabulated in Table 4. The computational estimation of 2D and 3D J-
integral with and without bimodularity is tabulate in Table 5 for all
contours with corresponding average value.
The methodology used for evaluating the 3D J eintegral has
been followed as a stress dependent elasticity problem with
bimodular characteristics. Fig. 15 (a) and (b) show that the value of
3D J-integral is significantly higher in bimodular stress field in
comparison to the unimodular stress field. Area integral due to
integrating in the thickness direction is significant which is not to
222 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225

Fig. 13. The Von-Mises stress plot for graphite having bimodular ratio 1.51 and subjected to mean peak load 94.28 N.

Fig. 14. Young's Modulus of elasticity plot for graphite having bimodular ratio 1.51 and subjected to mean peak load 94.28 N.

Fig. 15. Comparison of overall 3D J-integral with its constitutive line and area integral for (a) Unimodular specimen (b) Bimodular specimen. (Bimodular Elasticity ratio ¼ 1.51).

be neglected. The value of Line integral is equivalent to 2D Rice's J-
integral [30]. However, the 2D J-integral values are found to be
significantly higher in the both unimodular and bimodular stress
fields in comparison to corresponding 3D J-integral values. There-
fore, it seems, the 2D J-integral value overestimates the fracture
parameter for both the cases. This has also been ascertained from
experimental results as corroborated in Tables 4 and 5. This can be
concluded here that 3D J-integral is more accurate for quantifying
fracture behaviour. Still it remains to be established whether a
bimodular 3D J-integral or unimodular 3 D J-integral is a better
estimator for crack growth characterization. This has also been
ascertained from the evaluation of critical strain energy release rate
GIc as displayed in Fig. 15 and noted in Table 4 as 228.16 J/m2. Since
the material exhibits bimodular elastic behaviour, from Table 5, it
has also been observed that the corresponding bimodular 3D J
value (¼ 233.167 J/m2) is close to the GIc values. This indicates the
appropriateness and importance of bimodular 3D J-integral for
studying fracture behaviour of materials exhibiting different stress
strain behaviour in tension and compression.
Fig. 16 represents the plotting of GIc and JIc against the peak load
corresponding to SENB fracture experimentation. This plot repre-
sents the difference between experimentally evaluated values be-
tween two methods using the same ASTM standards. The authors
tried to find out the appropriate reason of this distinct behaviour of
almost equivalent quantities. The reason is that the method of
determining of KIc is through the peak load and model geometry, so
 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225 223

4. Conclusions

Fig. 16. Comparison of JIc (evaluated from experimental KIc) and GIc.
Table 4
JIc (evaluated from experimental KIc) and GIc
value for mean peak load estimated from
Fig. 16.
JIc GIc
J/m2 J/m2
184.28 228.16
that the method doesn't able to include the effect of bimodularity.
The value of JIc is calculated with the help of KIc value. But in the
case of determination of GIc, individual data points of load versus
displacement plot are involved, which includes the inherent
bimodularity. The critical energy release rate has been seen to be
more close towards the bimodular 3D J-value. Therefore, bimodular
consideration for the evaluation of energy release rate is justified in
comparison to determining stress intensity factor from peak load
data, where the later has been the prevailing practice for most cases
involving fracture dominant failures. Ignoring bimodularity in such
cases obviously might lead to catastrophic failure with susceptible
unreliable design adherences. Nevertheless, involving bimodularity
a design parameter makes a problem perturbation to stress
dependent elasticity phenomena with its inherent experimental
and computational complexity, though it might be unwise to ignore
the effect for high risk parts.
In the present work, the main focused area is about the four
main issues: (i) the evaluation of stress intensity factor from ex-
periments, (ii) the evaluation of critical energy release rate from
experiments, (iii) the prediction of mean peak load with Weibull
statistical distribution, (iv) evaluation of computational unim-
odular and bimodular 3D J-integral value and (v) comparison of
unimodular and bimodular results from experimental value.
Experimentations as per specific ASTM standards for tensile,
compressive and fracture specimens have been carried out to
investigate the influence of bimodularity on the crack growth
behaviour of graphite specimens. Weibull model has been devel-
oped for the randomized failure data set. The mean peak load so
obtained has been used in the finite element analysis of equivalent
numerical SENB specimens using program subroutines and COM-
SOL finite element programming software in a server based
computing environment. Following conclusions have been sum-
marized from the present investigation.
1. Bimodularity of graphite specimens have been evaluated by
conducting tensile and compression testing of twenty speci-
mens each as per ASTM standards. The bimodular ratio is found
to be 1.51 from mean normalizing of the data set. Significant
difference between JIc and GIc values have been observed from
the three-point bend SENB fracture experimentation of graphite
specimens.
2. The experimental evaluation of fracture toughness as per ASTM
D7779 standard only considers peak load and not the compli-
ance or the load displacement behaviour and hence the error is
accumulated when bimodularity is considered. However,
experimental evaluation of the critical strain energy release rate
considers the characteristic load versus displacement behaviour.
When material compliance exhibits different stress versus strain
behaviour in tension and compression the later is found to be
more appropriate to study the fracture behaviour of graphite
specimens. This has been reflected when the JIc calculated from
fracture toughness (from experiment) is compared with GIc
calculated from the load vs displacement plot, the later includes
the effect of compliance for the bimodular specimen. The same
has also been compared with existing literature data with an
assumed Young's modulus in Table 2(a). The evaluation of GIc as
per ASTM standard invokes the compliance plot which obvi-
ously includes the bimodular behaviour. With reference to

Table 5
Unimodular and bimodular 3D J-integral and 2D Rice's J-integral is tabulated for all twelve contour with average corresponding value.

Contour No. 2D Rice's J-value (Rice 1968) 3D J-value 2D J-value 3D J-value

(Unimodular) (Unimodular) (Bimodular) (Bimodular)
J/m2

J/m2 J/m2 J/m2 J/m2

contour 1 210.406 192.2617 259.772 236.0898

contour 2 212.9478 192.5205 259.2967 233.9932
contour 3 214.1847 192.5735 259.8547 233.8168
contour 4 214.268 192.1042 259.5517 233.6574
contour 5 212.9876 191.1078 257.9076 233.004
contour 6 209.8598 190.9858 254.2022 231.7182
contour 7 209.3302 190.8291 252.8956 231.0435
contour 8 208.7899 192.5331 251.7419 230.4144
contour 9 214.2231 192.6603 259.924 233.8705
contour 10 214.2591 192.1041 259.5141 233.7312
contour 11 213.1167 192.5791 258.3596 233.1376
contour 12 214.1994 191.3736 259.1634 233.5294
Avg value 212.381 191.9694 257.682 233.1672
224 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225
Table (2b), the differences found are significant to be ignored, if
the material behaviour is bimodular.
3. The mean peak load, Weibull modulus and coefficient deter-
mination has been evaluated using least square estimator
(LIN2), biased (MLE2-B) and unbiased (MLE2-U) maximum
likelihood estimator. Both two and three parameter Weibull
distribution has been carried out to find Weibull characterizing
parameter and mean peak load. MLE2-U estimator based mean
peak load has been used for the finite element analysis of single
edged cracked specimens as similar to the SENB experimental
specimens. The JIc and GIc values for the specimen tested are
found to be 184.28 J/m2 and 228.16 J/m2 respectively for the
characteristic mean peak load (94.28 N) evaluated from the
Weibull analysis.
4. Stress intensity factor and energy release rate is determined
from the adequate number of specimens tested experimentally
and the material mean critical stress intensity factor and mean
critical energy release rate is evaluated with the help of Weibull
distribution plot between peak load and cumulative probability
of failure. At the mean peak load corresponding KIc and JIc value
are the mean values which represent the properties of material.
5. Finite element analyses have been conducted to determine the
unimodular and bimodular 2D J-integral and 3D J-integral as
characterizing fracture parameter from the numerical simula-
tion. Analytical expressions have been derived for the bimodular
3D J-integral and the same has been incorporated through
subroutines to the finite element program. Path independence
of the contours have been shown for the twelve different con-
tours implying the validity and efficacy of the model to deter-
mine the crack tip stress field.
6. The 2D J-integral values are found to be significantly different
from 3D J-integral values. It has been observed that 3D J-integral
values including bimodularity are close to the critical strain
energy release rate GIc obtained for bimodular graphite speci-
mens. This has been reasoned due to the difference of unim-
odular and bimodular compliance as reflected in the evaluation
of unimodular and bimodular fracture parameters. Influence of
bimodularity is evident from the tabulated fracture parameters
and energy release rate plots.
The significant difference between the unimodular JIc and
bimodular GIc indicates that it will be unwise to ignore the effect in
high risk components. Therefore, load bearing structures should
involve the bimodularity parameter to lessen the unpredictability
of the strength failure and fracture of susceptible components
comprising of bimodular materials.
Acknowledgement
Both the authors wish to express their sincere gratitude for the
financial support provided by the grant-in aid of Department of
Atomic Energy, India scientific research grant under Grant No.
2011/36/62-BRNS.
References
[1] S.A. Ambartsumyan, Elasticity Theory of Different Modulus, 1986.
[2] S.A. Ambartsumyan, The axisymmetric problem of a circular cylindrical shell
made of material with different strength in tension and compression, izvestia,
Mekhanika 4 (1965) 1055e1067. http://oai.dtic.mil/oai/oai?
verb¼getRecord&metadataPrefix¼html&identifier¼AD0675312.
[3] S.A. Ambartsumyan, Equations for a plane problem of the multiresisting or
multi-modulus theory of elasticity, isvestia an armyaskoy SSR, Mekhanika 2
(1966).
[4] S.A. Ambartsumyan, Basic equations and relations in the theory of elasticity of
anisotropic bodies with different moduli in tension and compression,
Mekhanika Tverd. Tela Kiev. 3 (1969) 51e61.
[5] S.A. Ambartsumyan, A.A. Khachatryan, Some problems in the zero-moment
theory of shell made of materials with different moduli, SSR, Doklady 43
(1966) 198e204.
[6] A.A. Ambartsumyan, S.A. hachatryan, Elasticity for materials with different
resistance to tension and compression, Mekhanika Tverdogo Tela (in Russian)
2 (1966).
[7] S.A. Ambartsumyan, A.A. Khachatryan, A multi-modulus elasticity theory,
Mekhanika Tverd. Tela Kiev. 6 (1966) 64e67.
[8] C.W. Bert, J.N. Reddy, V.S. Reddy, W.C. Chao, Bending of thick rectangular
plates laminated of bimodulus composite materials, AIAA J. 19 (1981)
1342e1349, https://doi.org/10.2514/3.60068.
[9] C.W. Bert, Model for fibrous composites with different properties in tension
and compression, J. Eng. Mater. Technol. ASME 99 (1977) 344e349.
[10] C.W. Bert, F. Gordaninejad, Deflection of thick beams of multimodular mate-
rials, Int. J. Numer. Meth. Eng. 20 (1984) 479e503.
[11] C.W. Bert, J.N. Reddy, Mechanics of bimodular composite structures, Mech.
Compos. Mater. (1983) 323e337, https://doi.org/10.1016/B978-0-08-029384-
4.50027-4.
[12] https://www.gen-4.org/gif/jcms/c_9362/vhtr, (n.d.). https://www.gen-4.org/
gif/jcms/c_9362/vhtr.
[13] S.-H. Chi, Comparison of fracture toughness (KIC) and strain energy release
rate (G) of selected nuclear graphites, J. Nucl. Mater. (2016), https://doi.org/
10.1016/j.jnucmat.2016.04.040.
[14] N.N. Nemeth, R.L. Bratton, Statistical Models of Fracture Relevant to Nuclear-
grade Graphite: Review and Recommendations, 2011. http://ntrs.nasa.gov/
archive/nasa/casi.ntrs.nasa.gov/20110008737.pdf.
[15] G.E. Bacon, A method for determining the degree of orientation of graphite,
J. Appl. Chem. 6 (1956) 477e481, https://doi.org/10.1002/jctb.5010061101.
[16] G.M. Jenkins, Fracture in reactor graphite, J. Nucl. Mater. 5 (1962) 280e286,
https://doi.org/10.1016/0022-3115(62)90068-5.
[17] D.G. Harris, A.B. Miller, R.L. Craik, Control of anisotropy in the manufacture of
graphite, in: Symposium on Uranium and Graphite, Institute of Metals, Lon-
don, 1962, p. 67.
[18] K. Tsang, B. Marsden, Constitutive material model for the prediction of stress
in irradiated anisotropic graphite components, J. Nucl. Mater. 381 (2008)
129e136.
[19] B. Marsden, G. Hall, O. Wouters, J. Vreeling, J van der Laan, Dimensional and
material property changes to irradiated Gilsocarbon graphite irradiated be-
tween 650 and 750 C, J. Nucl. Mater. 381 (2008) 129e136.
[20] M. Srinivasan, On estimating the fracture probability of nuclear graphite
components, J. Nucl. Mater. 381 (2008) 185e198, https://doi.org/10.1016/
j.jnucmat.2008.07.029.
[21] S. Yeo, J. Yun, S. Kim, M.S. Cho, Y.W. Lee, Fabrication methods and anisotropic
properties of graphite matrix compacts for use in HTGR, J. Nucl. Mater. 499
(2018) 383e393, https://doi.org/10.1016/j.jnucmat.2017.11.055.
[22] T.D. Eto, M. Ishiyama, S. Nishiyama, Y. Oku, T. Burchell, Fracture toughness of a
fine-grained nuclear graphite by means of various methods, in: International
Symposium on Carbon, New Processing and New Applications, 1988,
pp. 170e173.
[23] S. Ishiyama, M. Eto, Fracture toughness of HTGR graphites, J. Soc. Mater. Sci.
Jpn 41 (1992) 475e481, https://doi.org/10.2472/jsms.41.475.
[24] G.D. Quinn, Weibull effective volumes and surfaces for cylindrical rods loaded
in flexure, J. Am. Ceram. Soc. 86 (2003) 475e479, https://doi.org/10.1111/
j.1151-2916.2003.tb03324.x.
[25] G.D. Quinn, Weibull strength scaling for standardized rectangular flexure
specimens, J. Am. Ceram. Soc. 10 (2003) 508e510.
[26] A. Bhushan, S.K. Panda, D. Khan, A. Ojha, K. Chattopadhyay, H.S. Kushwaha, I.
a. Khan, Weibull effective volumes, surfaces, and strength scaling for cylin-
drical flexure specimens having Bi-Modularity, ASTM J. Test. Eval. 44 (2016)
20150301, https://doi.org/10.1520/JTE20150301.
[27] B. Saint-Venant, in: Notes to Navier's Resume des lecons dela resistance des
corps solids, third ed., 1864. Paris.
[28] S. Timoshenko, in: Strength of Materials, Part 2, Advanced Theory and Prob-
lems, second ed., 1941. Van Nostrand, Princeton, NJ.
[29] A.A. Ambartsumyan and S.A. Khachatryan, Zero-moment theory for shells
made of materials with diverse tensile strength and compression strength, (in
Russian), AIAA Ref. No. A67-25569.
[30] J.R. Rice, A path independent integral and the approximate analysis of strain
concentration by notches and cracks, J. Appl. Mech. 35 (1968) 379e386,
https://doi.org/10.1115/1.3601206.
[31] F. Tabaddor, Constitutive equations for bimodulus elastic materials, AIAA 10
(1972) 516e518.
[32] F. Tabaddor, Analysis for beams made of bi-modulus elastic orthotropic ma-
terials, Fibre Sci. Technol. 9 (1976) 51e62.
[33] F. Tabaddor, Two-dimensional finite element analysis of bi-modulus mate-
rials, Fibre Sci. Technol. 14 (1981) 229e240.
[34] N. Kamiya, An energy method applied to large elastic deflection of a thin plate
of bimodulus material, J. Struct. Mech. 3 (1974) 317e329, https://doi.org/
10.1080/03601217408907270.
[35] N. Kamiya, Department of mechanical engineering, mie university, kamiha-
macho, tsu 514, Japan, Nucl. Eng. Des. 32 (1975) 351e357.
[36] R.M. Jones, D. a. R. Nelson, Theoretical-experimental correlation of material
models for nonlinear deformation of graphite, AIAA J. 14 (1976) 1427e1435,
https://doi.org/10.2514/3.61410.
[37] R.M. Jones, D. a. R. Nelson, Material models for nonlinear deformation of
 A. Bhushan, S.K. Panda / Journal of Nuclear Materials 503 (2018) 205e225
graphite, AIAA J. 14 (1976) 709e717, https://doi.org/10.2514/3.61410.
[38] R.M. Jones, Stress strain relations for materials with different moduli in ten-
sion and compression, AIAA 15 (1977) 16e23.
[39] B.C. Mitchell, J. Smart, S.L. Fok, B.J. Marsden, The mechanical testing of nuclear
graphite, J. Nucl. Mater. 322 (2003) 126e137, https://doi.org/10.1016/S0022-
3115(03)00322-2.
[40] J. Smart, B.C. Mitchell, S.L. Fok, B.J. Marsden, The effect of the threshold stress
on the determination of the Weibull parameters in probabilistic failure
analysis, Eng. Fract. Mech. 70 (2003) 2559e2567, https://doi.org/10.1016/
S0013-7944(03)00070-5.
[41] P. Ouagne, G.B. Neighbour, B. McEnaney, Crack growth resistance in nuclear
graphites, J. Phys. Appl. Phys. 35 (2002) 927e934. http://stacks.iop.org/0022-
3727/35/i¼9/a¼315.
[42] M.R. Joyce, T.J. Marrow, P. Mummery, B.J. Marsden, Observation of micro-
structure deformation and damage in nuclear graphite, Eng. Fract. Mech. 75
(2008) 3633e3645, https://doi.org/10.1016/j.engfracmech.2007.11.003.
[43] ASTM D7972-14, Standard test method for flexural strength of manufactured
carbon and graphite articles using three-point loading at room temperature 1,
ASTM Book. Stand. (2014) 1e5, https://doi.org/10.1520/D7972-14.2.
[44] C. Karthik, J. Kane, D.P. Butt, W.E. Windes, R. Ubic, In situ transmission elec-
tron microscopy of electron-beam induced damage process in nuclear grade
graphite, J. Nucl. Mater. 412 (2011) 321e326, https://doi.org/10.1016/
j.jnucmat.2011.03.024.
[45] T.H. Becker, T.J. Marrow, R.B. Tait, Damage, crack growth and fracture char-
acteristics of nuclear grade graphite using the Double Torsion technique,
J. Nucl. Mater. 414 (2011) 32e43, https://doi.org/10.1016/
j.jnucmat.2011.04.058.
[46] M.P. Hindley, M.N. Mitchell, D.C. Blaine, A.A. Groenwold, Observations in the
225
statistical analysis of NBG-18 nuclear graphite strength tests, J. Nucl. Mater.
420 (2012) 110e115, https://doi.org/10.1016/j.jnucmat.2011.09.013.
[47] M.P. Hindley, M.N. Mitchell, C. Erasmus, R. McMurtry, T.H. Becker, D.C. Blaine,
A.A. Groenwold, A numerical stress based approach for predicting failure in
NBG-18 nuclear graphite components with verification problems, J. Nucl.
Mater. 436 (2013) 175e184, https://doi.org/10.1016/j.jnucmat.2012.10.030.
[48] ASTM E1921-17a, Standard test method for determination of reference tem-
perature, to, for ferritic steels in the transition range, ASTM Book of Standards
(2017) 1e27, https://doi.org/10.1520/E1921-17A.2.
[49] ASTM D7779-11, Standard Test Method for Determination of Fracture
Toughness of Graphite at Ambient Temperature, 11, 2011, pp. 1e9, https://
doi.org/10.1520/D7779-11R15.the.
[50] ASTM C695-15, Standard test method for compressive strength of carbon and
graphite, ASTM Book. Stand. (2015) 1e3, https://doi.org/10.1520/C0695-15.2.
[51] ASTM C749 15, Standard test method for tensile stress-strain of carbon and
graphite 1, ASTM Book. Stand. (2015) 1e12, https://doi.org/10.1520/C0749-
15.2.
[52] ASTM E111-04, Standard test method for Young's modulus, tangent modulus,
and chord modulus, ASTM Book. Stand. (2010) 1e7, https://doi.org/10.1520/
E0111-04R10.
[53] ASTM C1239 13, Standard practice for reporting uniaxial strength data and
estimating Weibull distribution parameters for advanced ceramics, ASTM
Book. Stand. (2013) 1e18, https://doi.org/10.1520/C1239-13.Scope.
[54] W. Weibull, A statistical theory of the strength of materials, Proc. R.Swedish
Inst.Eng. Res 151 (1939) 1e45.
[55] W. Weibull, A statistical distribution function of wide applicability, J. Appl.
Mech. 18 (1951) 293e297 citeulike-article-id:8491543.