Nuclear Engineering and Design 263 (2013) 500–508

Contents lists available at SciVerse ScienceDirect

Nuclear Engineering and Design

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

Computational and experimental prediction of dust production in pebble bed

reactors—Part I
Maziar Rostamian a,∗ , Gannon Johnson a , Mie Hiruta a , Gabriel P. Potirniche a ,
Abderrafi M. Ougouag b , Joshua J. Cogliati b , Akira Tokuhiro a
a
Department of Mechanical Engineering, University of Idaho, 1776 Science Center Drive, Idaho Falls, ID 83401, USA
b
Idaho National Laboratory, 2525 N Fremont Avenue, Idaho Falls, ID 83401, USA

h i g h l i g h t s

• A nonlinear dimensionless wear coefficient is theoretically proposed.

• A material constant for the relation of asperity height and wear is introduced.
• A nonlinear modification of Archard wear formula is proposed.
• The graphite wear dust production in a typical pebble bed reactor is predicted.
• Experimental and computational wear results for graphite are presented.

a b s t r a c t
a r t i c l e i n f o
This paper describes the computational modeling and simulation, and experimental testing of graphite
Article history:
moderators in frictional contacts as anticipated in a pebble bed reactor. The potential of carbonaceous
Received 3 November 2012
Received in revised form 16 April 2013 particulate generation due to frictional contact at the surface of pebbles and the ensuing entrainment and
Accepted 19 April 2013 transport into the gas coolant are safety concerns at elevated temperatures under accident scenarios such
as air ingress in the high temperature gas-cooled reactor. The safety concerns are due to the documented
ability of carbonaceous particulates to adsorb fission products and transport them in the primary circuit
of the pebble bed reactor, thus potentially giving rise to a relevant source term under accident scenarios.
Here, a finite element approach is implemented to develop a nonlinear wear model in air environment.
In this model, material wear coefficient is related to the changes in asperity height during wear. The
present work reports a comparison between the finite element simulations and the experimental results
obtained using a custom-designed tribometer. The experimental and computational results are used to
estimate the quantity of nuclear grade graphite dust produced from a typical anticipated configuration.
In Part II, results from a helium environment at higher temperatures and pressures are experimentally
studied.
© 2013 Elsevier B.V. All rights reserved.

1. Introduction specific to Graphite-Moderated Gas Cooled Reactors (GM-GCR) is

The vast majority of today’s nuclear reactor designs, pre-
dominantly light water reactors, have fixed fuel and coolant
configurations. The pebble bed reactor (PBR), having neither of
these, uses nuclear fuel encased in a random configuration of spher-
ical fuel elements called ‘pebbles’ in a cylindrical ‘bed.’ While this
design is capable of online refueling and high output temperatures,
its design also creates specific modeling and design challenges for
safety and licensing (IAEA, 2003). Of these challenges, one that is
∗ Corresponding author. Tel.: +1 208 310 9555.
E-mail addresses: mrostamian@asme.org,
mzram 22@yahoo.com (M. Rostamian).
the phenomenon of carbonaceous particulate generation due to
frictional contact of graphitic components. Due to the number of
moving carbonaceous components (i.e. fuel elements or ‘pebbles’),
this is primarily a concern for pebble bed reactors for safety and
operation.
Safety concerns are focused around understanding transport
and deposition of fission and activated products, and how the dust
will be transported during a Depressurized Loss-Of-Flow Cool-
down accident (DLOFC). In particular, under DLOFC and similar
hypothetical accidents, the main issues are as follows: quantifica-
tion of the inventory of dust, dust re-suspension off of surfaces in
the primary circuit, dust departure from a breached primary circuit,
contamination levels associated with such an event (‘source term’),
and assessment of the additional risks posed by any energetic

0029-5493/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.nucengdes.2013.04.019
 M. Rostamian et al. / Nuclear Engineering and Design 263 (2013) 500–508
mechanisms of the graphite dust suspending into the surrounding
air (NUREG/CR-6944).
While there is limited data on these phenomena in the pub-
lished literature from past GM-GCRs and R&D, it is not sufficient
for the level of details expected under modern design review and
licensing requirements of a pebble bed by the United States Nuclear
Regulatory Commission (US-NRC). This is further complicated by
the fact that some materials in the past GM-GCRs are either not
used in current designs (e.g. Inconel 617 being replaced by modern
alloys), or available (e.g. nuclear graphite H-451) (Kissane, 2009).
A panel of experts ranked the phenomena as “medium” in the
Phenomena Identification Ranking Table (PIRT) and the knowledge
level as “low” (NUREG/CR-6944). The qualitative ratings in a PIRT
of “high,” “medium,” and “low” are used to identify the impor-
tance of expected phenomena and the level of understanding of
those phenomena. The US Nuclear Regulatory Commission Regu-
lation (NUREG) also identifies a need for graphite dust production
and tribological data in a helium environment, as an anticipated
function of temperature, pressure, and fluence.
In an effort to quantify the frictional wear of materials (fuel,
non-fuel) used in the pebble bed, a custom tribometer was devel-
oped to measure the wear of materials in helium atmosphere at
pressures and temperatures expected in a modern GM-GCR. This
paper describes the design, deployment of the device and the ini-
tial experimental data from several grades of nuclear graphite. Also,
in this paper the analytical linear Archard model is modified to
account for the wear of brittle materials. In an attempt, the wear
coefficient in this equation, which is conventionally assumed con-
stant, is derived as a function of the surface asperity height. The
finite element computational model of contacting fuel pebbles is
developed in ABAQUS.
2. The computational analysis
Wear in metals is commonly in the form of microcracks ini-
tiating under high local plastic strains. In the contact region, the
highest plastic strain is seen on the perimeter of the contact area.
This is where microcracks are likely to initiate. Existence of a con-
tinuous contact stress can easily cause to propagate the microcrack.
For plane stress, there are two major crack modes: mode I where a
tensile stress is normal to the plane of the crack, and mode II where
a shear stress acts parallel to the plane of the crack and perpendic-
ular to the crack front. Under the frictional contact condition, mode
I is the possible crack type. Microcracks that are formed can propa-
gate only if their stress intensity surpasses the critical crack stress
intensity as formulated by Irwin (1957). If this criterion is satisfied,
the cracks usually start to propagate perpendicular to the maxi-
mum principal stress (Pook, 2007). If the contact stress continues
to exist or if the segment under study goes through a high num-
ber of cycles, cracks on both sides of a 2D contact zone can meet
one another at a certain depth underneath the contact zone. This
mechanism produces chips, which leads to abrasive wear. Abrasion
can be a form of wear in brittle materials. However, at very small
inter-pebble loadings present in PBRs, this needs to be investigated.
Following a previous macro-level analysis (Rostamian et al., 2012),
and using eXtended Finite Element Method (XFEM), crack domains
were considered on the surface of two graphite pebbles in contact.
Considering the domain of applicability for the inter-pebble forces,
which is a range of 10–50 N (Cogliati and Ougouag, 2008; Rostamian
et al., 2012), crack initiations studies were performed. However, it
is revealed that the wear mechanism by plastic deformation and
crack initiation has an onset of 1000 N for IG-11 graphite pebbles
with a diameter of 3 cm. Therefore, it is assumed that the wear in
brittle materials such as graphite under smaller loadings is caused
by powder formation from the frictional contact between surface
501
asperities. This is also seen in the experimental results where after
60 m of frictional contact, only a small worn mass of approximately
1 ␮g was accumulated (Johnson, 2012). Therefore, an analytical
model has been established to consider the effect of asperities on
graphite wear.
Most brittle materials including graphite have a linear wear
behavior for small sliding distances (Blau, 1992). However, for most
real-life applications, the materials that are subject to frictional
contact are used over long time periods, which means they undergo
long sliding distances. In a previous work, it was attempted to
model a 2-m sliding wear (Rostamian et al., 2012). The results were
compared with Cogliati’s results (Cogliati and Ougouag, 2010) and
his reviews on Xiaowei’s results (Cogliati et al., 2011). However, it
is now attempted to model a much longer sliding length, so that
the nonlinear wear map (i.e. wear versus distance variation) gen-
erated from the experimental results can be predicted. In order to
capture this phenomenon, the linear Archard wear model has been
examined, modified and adapted.
2.1. The linear Archard wear model
The Archard wear equation is a linear model derived from
pin-on-disk (perpendicular contact) and pin-on-pin (tangential
contact) experiments by the French scientist, J.F. Archard, in the
1950s (Archard and Hirst, 1956). In this model, the wear volume
is related to the material properties and the sliding distance in the
following form
FL
V =K (1)
H
where K is the dimensionless wear coefficient or the proportional-
ity constant, which is a material parameter; F is the force normal to
the contact point; L the sliding distance; and H the material hard-
ness. It is seen that the wear volume is directly related to the normal
forces and the sliding distance. By performing finite element sim-
ulations at a macro-level, the Hertzian normal forces can be easily
determined (Johnson, 1987). As a result, the relationship between
the wear volume/mass and the sliding distance can be obtained
(Rostamian et al., 2010, 2011, 2012). However, this is only true for
small sliding distances in the case of brittle materials. In order to
account for the nonlinear wear rates or behavior of brittle mate-
rials, one needs to consider the material surface at a micro-level,
where asperities are considered as the real contact points. In the
following section, a micro-level study of the contact surface will
be considered and the role of asperity height in presented in the
newly introduced nonlinear version of the Archard model will be
explained.
2.2. Micro-level considerations
The height of asperities and the surface topology determine the
roughness of the material surface. Let us consider two segments
made of one material; one of a rough surface and one of a smooth
surface. It is obvious that the material of a rougher surface has a
higher wear volume after a certain distance of sliding wear. This is
also seen in the air test experiment results from the tribometer at
room temperature. The wear ring around the pebbles is seen to have
been smoothed after the pebbles have been in contact (see Section
3 and Fig. 6). Therefore, by a closer examination of the Archard wear
model, we strive to assess the dependence of wear volume on the
asperity height.
502 M. Rostamian et al. / Nuclear Engineering and Design 263 (2013) 500–508

2.3. A new wear correlation where ˛ = (hi H)/(Lmax P). Now, by introducing the following asper-
ity height power-law function, we strive to model the effect of
Based on the microlevel consideration above, we consider the surface smoothing:
volumetric wear rate (V̇ = dV/dt) form of the Archard equation:
n
(h∗) = R(h∗) (10)
F L̇
V̇ = K (2)
H where R is an ‘asperity model constant’ to be obtained by compar-
Introducing the asperity height into this equation, we approxi- ing the analytical model with the computational data and n is the
mate the volume by V = hA. Therefore, exponent of the power-law function, also to be determined through
a fit of the computational results. Substituting the asperity height
F L̇ function (h∗) into Eq. (9), we will have:
−ḣA = K (3)
H ⎧ dh∗
where A is the contact area. The negative sign denotes that the time ⎪
⎪ −˛ = K, h = hi (11a)
⎨ dL∗
rate change of asperity height is negative, representing a decrease ∗
dh
in asperity height with time. It should be noted that the contact ⎪
⎪ −˛ = K (h∗ ) = KR · (h∗ ) , h < hi (11b)
n

area can vary based on the component shape and the contact con-
⎩ dL∗  
K
figurations. In order to generalize this equation for any contact area
and component shape, we consider the contact stresses instead of n
where K  = KR(h∗ ) is a varying dimensionless wear coefficient,
normal forces. Then, Eq. (3) takes the form:
which depends on the variation of asperity height with wear length
P L̇ and on the material properties. By rearranging and integrating, the
−ḣ = K (4) varying wear coefficient, K  , can be derived as a function of the
H
sliding length, L*, and the constants R and n. There follows,
where P is Hertzian contact stress. By performing computational
simulations in ABAQUS, contact stresses can be easily obtained at
1/(1−n)
 KR
each node in the mesh. In Eq. (4), L̇ is sliding distance rate or relative K = . (12)
velocity at contact. However, the time rate of change in the asperity [(1 − n)˛L∗ ]n
height, ḣ, is unknown. To eliminate this unknown, we consider the
The wear coefficient thus derived depends on two parameters:
following:
the asperity model constant, R, and the power law coefficient n.
h = h(t) = h(L(t)) (5) These material constants are determined through a fitting process
described here. The finite elements computational model pre-
which denotes that the dependence of asperity height on time can sented in Section 2.4 allows the determination of asperity height
be formulated as the dependence of asperity height on wear length. as a function of wear length. The FEA model is used repeatedly to
Here, the gas adsorption effects on graphite surface are not consid- obtain a function describing the height, h, of a given asperity versus
ered. In order to find ḣ, the partial derivative of h with respect to the wear length, L, as it is varied between 0.0 and 2.0 m. The deriva-
time is used: tive of that function is proportional to K  . Knowing the computed
(L, K  ) pairs and introducing the L values into Eq. (12) for K  , “model”
dh dh dL dh ḣ dh
= · ⇒ ḣ = · L̇ ⇒ = (6) values for K  are obtained for various values of the parameters R and
dt dL dt dL L̇ dL
n. This process allows the formulation of a fitting problem for the
Therefore, R and n parameters, based on minimizing the difference between
the computed K  and the “model” K  . The values of R and n that are
H dh retained in this work are the first pair found that results in a max-
− =K (7)
P dL imum absolute error below 10−6 for K  . The parameters obtained
This is still a linear equation for the rate of change of the asper- in this way from computational model data up to 2.0 m of wear
ity height. However, it can now be argued that the asperity height length are then used for predicting the wear performance for up
which decreases with the sliding length can start to influence the to 1200 m of wear length. The analytical model predicts experi-
material constant K. This material property is commonly consid- mental data with remarkable accuracy and fidelity as detailed later
ered as a constant when the Archard equation is implemented for through comparison with the experimental results. All the fitting
wear calculations. In this work, we introduce a function, which take work described here assumed the properties of graphite IG-11.
into account the effect of asperity height variations. This function
incorporates the effect of asperity height decrease into the linear
2.4. The computational analysis
Archard wear model for higher sliding distances. When the asper-
ity height decreases from its initial height, the surface becomes
The finite element software ABAQUS v6.11 has been used to con-
smoother, which indicates that the wear rate should decrease.
duct simulations of the micromechanics of wear (ABAQUS user’s
Before introducing the asperity height function, we consider the
manual). Primarily, 3D spherical asperities were considered to
following normalized parameters to simplify the equations.
be modeled using the implicit module of ABAQUS known as the
L h ‘ABAQUS Standard’. However, given the demanding nature of con-
L∗ = , h∗ = (8) tact conditions, the nonlinear geometry of the asperities, severe
Lmax hi
deformations, and most importantly, the overlapping of asperi-
where Lmax is the maximum sliding length and hi is the average ties, the implicit module of ABAQUS faced convergence issues. An
asperity height of the component unworn surface. Therefore, Eq. alternative was taken by applying the same methodology in the
(7) takes the form explicit solver of ABAQUS known as the ‘ABAQUS Explicit’, which
dh∗ is very robust in handling contact problems. This module however
−˛ =K (9) demands higher run times. As a result, it was decided to switch
dL∗
from spherical to cylindrical asperities.
 M. Rostamian et al. / Nuclear Engineering and Design 263 (2013) 500–508 503

occurrence of hourglass deformations in an analysis can invalidate

the results and should always be minimized. Therefore, hour-
glass control is used to overcome this deficiency. This formulation
provides improved coarse mesh accuracy with slightly higher com-
putational cost and performs better at high strain levels. Hourglass
control is an embedded feature in ABAQUS v6.11.
Plane strain is simulated in this 3D model by constraining
the front and back planar faces of the asperity as symme-
try (i.e. uz = 0, urx = ury = 0). In this way, the model behaves
as a plane strain slice. A prescribed velocity was primarily
used for the movement of the upper asperities, and the lower
asperity was kept fixed as its base. Hence, to simulate the move-
ment of both pebble surfaces, the relative tangential velocity is
applied to the upper asperity. In a later attempt, it was found
out that prescribing velocities to both upper and lower asper-
ities in opposite directions provides more realistic results for
the two asperities. Therefore, the actual tangential velocity of
pebbles at contact was applied to upper and lower asperity
bases.

2.4.2. Mass scaling

Modeling using an explicit procedure can take a considerable
Fig. 1. A meshed asperity of radius 3 ␮m and thickness 0.04 ␮m.
amount of time for the simulation to converge. This is due to the
fact that the minimum stable time increment is small. Using appro-
2.4.1. Finite element model priate mass scaling, the computational efficiency is improved while
Fig. 1 shows the finite element mesh used for the cylindrical retaining the necessary degree of accuracy required for a particular
asperity model. This asperity model was made of linear hex- class of problems.
ahedral elements (C3D8R). ‘R’ stands for ‘reduced-integration’, Mass scaling was introduced into the material model in mul-
which uses fewer number of Gaussian points when solving the tiple steps. This means that in every step, an artificial mass
integral. The advantage of the reduced integration elements was added and the simulations were performed. After assess-
is that the strains and stresses are calculated at the loca- ing the results, if inertial effects were not seen as compared
tions that provide optimal accuracy; the so-called Barlow points with the results of no artificial mass, then some more mass
(Barlow, 1976). A second advantage is that the reduced number was introduced. Otherwise, mass scaling would not be contin-
of integration points decreases CPU time and storage require- ued.
ments.
In the case of cylindrical asperities which is considered 2.4.3. Material model
in this study, it would be common sense to conduct the The material properties for nuclear grade graphite IG-11 were
simulation in 2D considering shell elements. However, since chosen for the asperity model. Here we have used a damage
element removal is sought in these simulations, the ele- model and have included damage evolution to visualize the degra-
ments which are originally within the interior of the asperity dation of the fully damaged elements. Material failure refers to
can start to interact with the opposing asperity as a con- the complete loss of load-carrying capacity, which results from
tact surface. This feature is only avalibale in ‘general contact
algorithm’ in ABAQUS, which requires the use of 3D solid
elements. Thus, a 3D model of cylindrical asperities was con-
structed.
Despite being robust for large deformations and saving exten-
sive amounts of run time, the reduced-integration solid and shell
elements used in ABAQUS are prone to zero-energy modes. These
modes, commonly known as hourglass modes, are oscillatory in
nature and result in mathematical states that are physically unfea-
sible. They typically have no stiffness and give a zigzag appearance
to the mesh known as hourglass deformations as seen in Fig. 2. The

Fig. 2. A schematic of hourglass deformation. Fig. 3. Elastic–plastic behavior with hardening for nuclear graphite IG-11.
504 M. Rostamian et al. / Nuclear Engineering and Design 263 (2013) 500–508

progressive degradation of the material stiffness. An undam-

aged constitutive behavior of the material (i.e. elastic-plastic with
hardening) is determined (Yokoyama et al., 2008) as seen in
Fig. 3.
Then, a damage initiation criterion is determined (point A). This
is followed by damage evolution (paths A–B) and finally the choice
of element removal (point B).
In the absence of any experimental data for the damage
softening part of the curve (A–B), a simple linear damage evo-
lution law was used to degrade the material stiffness to zero
where the equivalent total strain reaches 0.05 (point B in
Fig. 3). When an element reaches point B on the curve, its stiff-
ness has fully degraded, and the element is removed from the
mesh.

2.4.4. The finite element analysis

The finite element analysis of asperity contact in ABAQUS
Explicit revealed that asperities undergo very high deforma-
tions. As the material reaches a total strain of 0.05, the
elements are removed and the degradation leads to mate-
rial removal from the asperity. In Fig. 4, the degradation
and material removal from asperities of 3 ␮m high are pre-
sented.
The height of the worn asperity from the first simulation was
used to perform the next simulation on this asperity after one full
rotation of the graphite pebble. This process continued to derive the
right hand side of Eq. (11b), which is needed to derive the ‘asperity
model constant’, R.

3. The experiment

The basic sample design conceived for testing to mimic the

geometry of the pebble bed with the apparatus is a graphite
disk machined into 2.54 cm tall disks with a 3 cm outer contour
(Johnson, 2012).

3.1. The experimental apparatus

The test data used in this paper is at room temperature and

atmospheric pressure. This is because the material properties for
nuclear graphite are only available at room temperature. Also, there
is no test data for graphite matrix or graphite at the tempera-
tures and pressures expected in a pebble bed. As a result, a custom
test apparatus was designed. In the following section, the exper-
imental setup is explained in brief. The motion requirements for
rolling and sliding between two spherical objects led to the selec-
tion of a twin disk tribometer. Because no off-the-shelf twin-disk
tribometer could operate at the temperature and pressure require-
ments of PBR, a custom apparatus had to be designed. However,
only the results at room temperature and atmospheric pressure
are reported in the present paper. High temperature and pres-
sure results are presented in a parallel but different work (Part
II). The samples were also measured before and after the experi-
ments on an analytical balance for a separate method to measure
the average wear rate of the samples for comparison. A schematic
of the tribometer with the pebbles in contact is presented in
Fig. 5.
The Scanning Electron Microscopy (SEM) images (Fig. 6) from
the graphite samples, after an approximate sliding distance of
1000 m show a wear scar of 1.7 mm wide with a smooth and
clean surface. This is while outside of wear scar, a rough sur-
Fig. 4. Equivalent plastic strain contour: the degradation and material removal pro-
face from the fabrication process is observed. This indicates cess from asperities at: (a) t = 0.000325 s, (b) t = 0.0013 s, (c) t = 0.002275 s and (d)
that the asperities were polished and leveled by the rotating t = 0.00325 s for a tangential speed of 0.08 m/s.
motion.
 M. Rostamian et al. / Nuclear Engineering and Design 263 (2013) 500–508 505

Fig. 5. Tribometer (left) and its schematic (right) with pebbles in contact and the modular analytical balance.

4. Results and discussion

The wear mass results from the tribometer are presented in

Fig. 7. The wear mass was recorded every 200 m. It is clearly seen
that the primary wear rates are reduced as the sliding length
increases. The wear coefficient obtained from the analytical model
is compared in Fig. 8a to the computational results, and in Fig. 8b
to the experimental data. The analytical model exponent and ‘the
asperity model constant’ were derived by fitting with the compu-
tational results. The model thus obtained is then compared with
the experimental data for verification purposes. The exponent, n, is
computed to be −1.38 and the dimensionless microlevel asperity
model material constant, R, is 2.5 × 10−4 .
It is clearly observed in Fig. 8b that the proposed analytical
model, which stems from the computational data, predicts a higher
overall wear coefficient than the actual wear coefficient derived
from experimental data. This is because the natural lubricating
properties of graphite surface have not been considered (Czichos,
1978). As explained earlier, when two graphite surfaces are in con-
tact, powder formation is the main form of wear. The powder is then
capable of filling in the surface valleys, which causes the surface

Fig. 6. The SEM images of graphite sample (a) wear scar zone (b) interface between
the scar (lower) and rough zone (upper). Fig. 7. Wear mass loss from the tribometer air-tests for ∼1100 m.
506 M. Rostamian et al. / Nuclear Engineering and Design 263 (2013) 500–508
Fig. 8. Comparison of analytical K with (a) computational results and (b) experi-
mental data with respect to the dimensionless sliding length, L*.
to become smoother than what is incorporated into our ABAQUS
model. A plausible explanation is that the uneven distribution of
asperity height on an actual surface could cause a different non-
linear wear coefficient than what is predicted here. The minimum
(final) asperity height, as predicted by this model, is 12.45% of the
initial average asperity height. This indeed indicates that the sur-
face has become smoother and thus the wear coefficient smaller.
Using the new varying wear coefficient, the wear results, which
were predicted before based on a constant coefficient, can now be
rectified. As seen in Fig. 9a, the initial wear rate starts to decrease
by wear distance and the cumulative wear is extremely small for
air-tests.
The predicted wear mass using the proposed analytical model
is compared to the experimental results in Fig. 9b. As anticipated,
due to the difference between the predicted and actual dimension-
less wear coefficients, the predicted wear mass is approximately
20% higher than the actual wear mass. It is clearly observed
in Fig. 9b that the proposed wear model shows a significant
improvement in comparison to the Archard wear equation. The
error bars indicated on the graph in Fig. 9b are from the uncer-
tainty associated with the scale with which the experimental
data was recorded. This indicates that the analytical model has
predicted reliable results that fit in within the uncertainty lim-
its.
Fig. 9. Instantaneous and cumulative wear mass for one graphite pebble with
respect to wear length, L.
The varying wear rate of graphite is shown in Fig. 10.
It is seen that the wear rate decreases as graphite asper-
ity height decreases with increasing wear length. The reported
wear rate in (Johnson, 2012) was a constant 8.45 × 10−8 . This
value is the same as the initial rate predicted in the present
study. However, in the present study, the rate is shown to be
varying over time. It must be noted that the present results
are for air test at room temperature and atmospheric pres-
sure.
Using the dimensionless wear coefficient K  obtained in this
work, it is attempted to derive the wear mass for one pebble
through 2.06 m of HTR-10 wear length as in (Rostamian et al., 2012).
By performing wear calculations as carried out by Cogliati et al.
(2011) based on the air-test results of Xiaowei et al. (2005a), the
wear mass produced by inter-pebble forces (Cogliati and Ougouag,
2008) is calculated. Estimates for HTR-10 and AVR are reported
using the present model. The wear mass for one pebble is esti-
mated to be 0.235 mg. For HTR-10 with an inventory of 27,000
pebbles and a pebble flow rate of 125 pebble per day, the wear
mass is predicted to be 5.36 g/year. As explained by Rostamian
et al. (2012), the estimated wear mass for AVR would then be
63.55 g/year.
 M. Rostamian et al. / Nuclear Engineering and Design 263 (2013) 500–508
Fig. 10. Wear rate for one graphite pebble with respect to wear length, L.
In Part II of this paper, helium-test experimental data at higher
temperatures and pressures will be reported. It is known that
higher temperature and pressure have significant impacts on the
wear properties of graphite. These effects will be discussed in
Part II.
5. Conclusions
In the present work, we investigated the generation of car-
bonaceous particulate due to frictional contact at the surface of
graphite pebbles. Such particulates can give rise to safety issues at
elevated temperatures, under accident scenarios for a high temper-
ature gas-cooled reactor. Specifically, the adsorbed fission products
onto the ‘graphite dust’ represent a source-term concern. Here, a
micro-level approach is undertaken to investigate the microme-
chanics of wear of nuclear graphite IG-11. The effect of the height
of surface asperities is introduced into the wear formula, and
a correlation between dimensionless wear coefficient and sur-
face asperity height is obtained. We learned that the likelihood
of chip/particulate formation that is the major source of wear in
metals is not dominant for porous, carbon-chain materials such as
nuclear grade graphite.
The constants in the proposed analytical wear model are
obtained by fitting with the computational results. Then, the model
is used to predict experimental data. In this way, the accuracy of
the proposed model is assessed. The computational results revealed
that there is approximately 20% difference between the results of
507
obtained from the analytical model and the experimental data.
This is attributed to the lack of a surface-level lubrication com-
ponent in the model, thought to be induced by gas adsorption
as well as relocation of surface-adhering graphite particulates. It
is known that graphite acts as a good lubricant. Thus it could
be surmised that when particulates are formed due to wear,
they primarily relocate in the valleys between asperities. This
increases the smoothness of the surfaces, which in turn decrease
the wear coefficient. The lamellar structure of graphite facilitates
the establishment of smooth surfaces between sliding contacts,
and the presence of adsorbed gases may further improve the
quality of these surfaces. Since these artifacts are not considered
in the finite element simulations, the predicted wear coefficient
is slightly larger than that attributed to the experimental data.
However, it is seen that the proposed model is capable of repre-
senting the nonlinear wear rates as seen in experimental results
and that it falls within the limits of the experimental data uncer-
tainty.
The proposed model and the experimental results show that
the wear mass for air environment at room temperature and
atmospheric pressure is small and on the order of grams per
year for a reactor. Therefore, it is predicted that the presence of
helium environment at higher temperature and pressures, actu-
ally contributes to increased wear. It was shown by Xiaowei
et al. (2005b) that the wear rate of graphite at elevated tem-
peratures is much higher than that at room temperature. High
temperature/pressure tests in helium gas were also tested using
the custom-designed tribometer. The results of these studies
are reported in Part II as a continuation of the work presented
here.
Acknowledgement
We express our gratitude to DOE, under NEUP09-151, “The
Experimental Study and Computational Simulations of Key Pebble
Bed Thermomechanics Issues for Design and Safety” for providing
financial support for this study.
References
ABAQUS Finite Element Analysis, SIMULIA, Dassault Systèmes, http://www.simulia.
com/support/documentation.html (visited May 2011–July 2012).
Archard, J.F., Hirst, W., 1956. The wear of materials under unlubricated conditions.
Proc. R. Soc. A 236, 397–410.
Barlow, J., 1976. Optimal stress locations in finite element models. Int. J. Numer.
Methods Eng. 10, 243–251.
Blau, P.J., 1992. ASM Handbook. Vol. 18: Friction, Lubrication, and Wear Technology.
ASM International, Niagara Falls, NY, USA.
Cogliati, J.J., Ougouag, A., 2008. Pebble bed reactor dust production model. In:
Proceedings of the 4th International Topical Meeting on High Temperature Reac-
tor Technology, Washington, DC, USA.
Cogliati, J.J., Ougouag, A.M., 2010. Dust Production Model for HTR-10. Idaho National
Laboratory, Report to Department of Energy.
Cogliati, J.J., Ougouag, A.M., Ortensi, J., 2011. Survey of dust production in pebble
bed reactor cores. Nucl. Eng. Des. 241 (6), 2364–2369.
Czichos, H., 1978. Tribology: A System Approach to the Science and Tech-
nology of Friction, Lubrication and Wear. Elsevier Scientific Publishing
Company, Bedford, NS, Canada.
IAEA, 2003. Evaluation of High Temperature Gas Cooled Reactor Performance:
Benchmark Analysis Related to Initial Testing of the HTTR and HTR-10,
IAEATECDOC-1382.
Irwin, G., 1957. Analysis of stresses and strains near the end of a crack traversing a
plate. J. Appl. Mech. 24, 361–364.
Johnson, K.L., 1987. Contact Mechanics. Cambridge University Press, Fairford, GLO,
UK.
Johnson, G., 2012. Experimental study of graphite–graphite and graphite–steel wear
in spherical contact in a pressurized inert atmosphere at elevated temperatures.
University of Idaho, Idaho Falls, Idaho (Master’s thesis).
Kissane, M.P., 2009. A review of radionuclide in the primary system of a very-high-
temperaure reactor. Nucl. Eng. Des. 239, 2076–3091.
508 M. Rostamian et al. / Nuclear Engineering and Design 263 (2013) 500–508
Pook, L.P., 2007. Metal Fatigue, Solid Mechanics and its Applications, vol. 145.
Springer, Guernsey, GY, UK.
Rostamian, M., Arifeen, Sh., Potirniche, P.G., Tokuhiro, A., 2010. Initial anal-
ysis of pebble contact in pebble bed reactors. Trans. Am. Nucl. Soc. 103,
1028–1031.
Rostamian, M., Arifeen, Sh., Potirniche, P.G., Tokuhiro, A., 2011. Initial pre-
diction of dust production in pebble bed reactors. J. Mech. Sci. 2,
189–195.
Rostamian, M., Potirniche, P.G., Cogliati, J.J., Ougouag, A.M., Tokuhiro, A., 2012. Com-
putational prediction of dust production in pebble bed reactors. Nucl. Eng. Des.
243, 33–40.
U.S. Nuclear Regulatory Commission Regulation, 2008. Next Generation Nuclear
Plant Phenomena Identification and Ranking Tables (PIRTs), NUREG/CR-6944
(1–3). Fission-Product Transport and Dose PIRTs.
Xiaowei, L., Suyaun, Y., Zhen-sheng, Z., Shu-yan, H., 2005a. Estimation of graphite
dust quantity and size distribution of graphite particle in HTR-10. Nucl. Power
Eng. 26, 0258–0926.
Xiaowei, L., Suyuan, Y., Xuanyu, S., Shuyan, H., 2005b. Temperature effect on IG-11
graphite wear performance. Nucl. Eng. Des. 235, 2261–2274.
Yokoyama, T., Nakai, K., Futakawa, M., 2008. Compressive stress–strain character-
istics of nuclear-grade graphite IG-11: effect of specimen size and strain rate. J.
Jpn. Soc. Nucl. 7 (1), 66–73.