INL/CON-23-71179-Revision-0

FINITE ELEMENT
MODEL MESH
REFINEMENT EFFECTS
ON QUALIFICATION OF
NUCLEAR GRADE
GRAPHITE
COMPONENTS
July 2023

Andrea L Mack, William E Windes, Joseph Bass, William M Hoffman

INL is a U.S. Department of Energy National Laboratory operated by Battelle Energy Alliance, LLC
 DISCLAIMER
This information was prepared as an account of work sponsored by an
agency of the U.S. Government. Neither the U.S. Government nor any
agency thereof, nor any of their employees, makes any warranty, expressed
or implied, or assumes any legal liability or responsibility for the accuracy,
completeness, or usefulness, of any information, apparatus, product, or
process disclosed, or represents that its use would not infringe privately
owned rights. References herein to any specific commercial product,
process, or service by trade name, trade mark, manufacturer, or otherwise,
does not necessarily constitute or imply its endorsement, recommendation,
or favoring by the U.S. Government or any agency thereof. The views and
opinions of authors expressed herein do not necessarily state or reflect
those of the U.S. Government or any agency thereof.
 INL/CON-23-71179-Revision-0

FINITE ELEMENT MODEL MESH REFINEMENT

EFFECTS ON QUALIFICATION OF NUCLEAR GRADE
GRAPHITE COMPONENTS

Andrea L Mack, William E Windes, Joseph Bass, William M Hoffman

July 2023

Idaho National Laboratory

Idaho Falls, Idaho 83415

[Link]

Prepared for the

U.S. Department of Energy
Under DOE Idaho Operations Office
Contract DE-AC07-05ID14517
 Proceedings of the ASME 2023
Pressure Vessels & Piping Conference
PVP2023
July 16-21, 2023, Atlanta, Georgia, USA

PVP2023-107369

FINITE ELEMENT MODEL MESH REFINEMENT EFFECTS ON QUALIFICATION OF

NUCLEAR GRADE GRAPHITE COMPONENTS

Andrea Mack1, William Hoffman1, Joseph Bass1, William Windes1

1
Idaho National Laboratory, Idaho Falls, ID

ABSTRACT Keywords: ASME, full assessment, simplified assessment,

The American Society of Mechanical Engineers (ASME) qualification, nuclear, graphite, Weibull, probability, failure,
provides the full and simplified design-by-analysis probabilistic finite element analysis, convergence, strength, stress
assessments for determining acceptance of nuclear grade
graphite core components. The assessments can be NOMENCLATURE
characterized by three parts: (1) a component stress distribution, α 3-parameter Weibull shape parameter
often determined by a finite element (FE) model; (2) a Weibull β 3-parameter Weibull scale parameter
probability density function (pdf) that characterizes the 𝜇𝜇 3-parameter Weibull threshold parameter
experimental tensile strength distribution; and (3) the post- 𝑉𝑉𝑚𝑚 minimum link volume
processor, which combines the FE model and the Weibull Δ relative stress range parameter
strength distribution in accordance with the full and simplified 𝑅𝑅𝑡𝑡𝑡𝑡 ratio of the flexural mean strength to the tensile mean strength
assessments to determine component acceptance. It is known 𝐶𝐶𝑚𝑚 combined membrane stress
that the level of mesh refinement in FE models can affect the ASME American Society of Mechanical Engineers
modeled component’s calculated stress distribution. Depending BPVC Boiler and Pressure Vessel Code
on the component geometry, the stress distribution may converge FE Finite Element
with sufficient refinement. It was previously unknown whether LB Lower Bound
the acceptance decision resulting from the full and simplified MLE Maximum Likelihood Estimate
assessments might change even with sufficient mesh refinement. POF Probability of Failure
This study explores that question using experimental strength POS Probability of Survival
results for a dog-bone geometry for two graphite grades, IG-110 SRC Structural Reliability Class
and PCEA. The simplified assessment has two criteria that must pdf probability density function
be met, the first limits the combined membrane stress by the cdf cumulative density function
allowable stress and the second limits the peak equivalent stress
by the allowable stress scaled by the ratio of flexural to tensile 1. INTRODUCTION
strength. In the application of the simplified assessment, The next generation of nuclear reactors include High
convergence of the peak equivalent stress required extreme mesh Temperature and Very High Temperature Reactor (HTR
refinement, however, the acceptance decision was not affected. and VHTR) designs [6]. These advanced reactor designs
It is hypothesized that more complex geometries with stress utilize thermally stable graphite as core internal
concentrations may present mesh refinement effects on the components [7] using an inert coolant, composed of either
simplified assessment acceptance decision. Mesh refinement did gas or molten salt. The American Society of Mechanical
affect the acceptance decision in the full assessment for the Engineers (ASME) Boiler and Pressure Vessel Code
applied pressure loadings in this study. This work suggests (BPVC) Section III, Division 5 specifies Rules for
component stress distribution convergence is not a sufficient Construction of Nuclear Facility Components for High
criteria for POF convergence in the full assessment and that Temperature Reactors [3]. Subsection HH Subpart A,
mesh refinement should continue until the POF has converged, Article 3000 (HHA-3000) provides design guidelines for
especially where the resulting POF is bordering the SRC class SN nonmetallic core components, specifically
acceptable POF limit. graphite components.

1 © 2023 by ASME
 Graphite is a quasi-brittle material with inherent pore convergence is met before mesh convergence in most cases
and flaw distributions [8,9]. Thus, HHA-3000 implements [4], but details are not explained. The studies in [4] only
probabilistic methods to set acceptance limits through two represent results for NBG-18 and the 2019 Code version.
design-by-analysis alternatives for qualification, the
simplified and full assessments. The simplified assessment 2. MATERIALS AND METHODS
uses a 2-parameter Weibull distribution to describe the
tensile strength distribution of the given graphite grade to 2.1 ASME BPVC HHA Methods
determine an allowable stress. Two requirements must be
met to satisfy the simplified assessment, one based on the There are five loading groups for which acceptance
component’s combined membrane stress (𝐶𝐶𝑚𝑚 ) and one limits are applied: Design and Service Levels A, B, C, and
based on the component’s peak equivalent stress, often D. The limits for each of the loading groups are dependent
found using finite element (FE) analysis. The full on the structural reliability class (SRC) of the component.
assessment uses a 3-parameter Weibull distribution and This work is limited to safety class components, SRC-1,
groups the elements throughout the component based on which are associated with a POF limit of 10−4 and Design
the magnitude of the equivalent stress. The output is a and Service Levels A, B and C. Figure HHA-3221-1 in the
component probability of failure (POF), which is compared ASME Code [3] outlines the criteria for graphite core
to a POF limit. Design-by-test is a third alternative but is component acceptance.
not discussed in this paper.
IG-110 and PCEA graphite grades with the dog-bone The assessments can be characterized by three parts: (1)
mechanical test geometry were used in this study to a FE model of the component; (2) a Weibull probability
investigate the effect of mesh refinement on the simplified density function (pdf) that characterizes the experimental
and full assessments. This paper explains the full and tensile strength distribution; and (3) the post-processor,
simplified assessment methodologies and applies those which combines the FE model and the Weibull strength
methods to dog-bone tensile test data fabricated from distribution in accordance with the full and simplified
current nuclear graphite grades IG-110 and PCEA. Finite assessments to determine component acceptance. The
element modeling is performed with five mesh refinements. simplified and full assessments are briefly described in the
The full assessment implements a step that groups mesh following subsections for immediate reference.
elements. The grouping criteria changed between the 2019
and 2021 Code versions, therefore this study is applied 2.1.1 Weibull Distribution
based on both Code versions. Results and implications of Both full and simplified assessments require
mesh refinement on component acceptance per HHA-3000 experimental tensile strength data and component FE
is discussed, as well as other considerations. equivalent tensile stresses as inputs. Both full and simplified
assessments make use of the Weibull pdf, denoted 𝑓𝑓(𝑥𝑥), to
1.1 Background characterize the grade’s random tensile strength distribution.
Neither the full nor simplified graphite component The pdf maps strengths to probabilities associated with
assessments of the ASME BPVC [3] specify FE modeling failure. The cumulative density function (cdf) is the integral
mesh refinement guidelines. It is common practice in FE of a pdf, and as its name implies, it provides the probability
modeling to continue mesh refinement until stress of failing at or below a given strength. The 3-parameter
distribution convergence has been obtained. While a Weibull cdf is defined by the shape (𝛼𝛼), scale (𝛽𝛽) and
designer might demonstrate stress convergence in their threshold (𝜇𝜇) parameters. Notation choice corresponds to
application, the question was opened as to whether stress that found in HHA-II-3200 of [3]. The 3-parameter Weibull
convergence corresponds to: (1) convergence in the peak cdf is given below.
equivalent stress used in the simplified assessment (HHA-
𝑥𝑥−𝜇𝜇 𝛼𝛼
II-3300); and (2) convergence in the POF found using the −� �
𝐹𝐹(𝑥𝑥|𝛼𝛼, 𝛽𝛽, 𝜇𝜇) = 1 − 𝑒𝑒 𝛽𝛽 (1)
methodology specified within the full assessment, sub-
article HHA-3217. Since no mesh refinement guidelines are
provided, it is necessary to understand whether refinement
The Code utilizes one-sided 95% lower bounds for the
beyond a typical converged stress distribution is necessary
shape and scale parameters. New notation is not introduced
to achieve consistent acceptance decisions.
for lower bounds in this paper. When the notation is
representative of lower bounds, it is indicated in text.
1.2 Related Work
Hindley performed much of the foundational work for
the full assessment implemented in HHA-3217. In his paper 2.1.2 Simplified Assessment
[4], he discusses having done some mesh refinement studies The simplified assessment is primarily laid out in
on POF convergence. In most cases, it appears Hindley used sections HHA-II-3100 and HHA-II-3300 and Figure HHA-
two or three mesh refinements. Hindley claims POF II-3221-1 in [3]. It uses the 2-parameter Weibull

2 © 2023 by ASME
distribution, which is found by setting the threshold 2.2 Code Assessment Assumptions
parameter to 0 in Equation 1. This makes strengths close to This study set out to apply the Code rules as they are
0 have a non-zero probability of failure, though still quite written in the 2019 and 2021 Code versions. Work presented
small. 𝐹𝐹(𝑥𝑥|𝛼𝛼, 𝛽𝛽, 𝜇𝜇) is set to the SRC-1 POF limit, at the SMiRT conference in 2015 [10] showed that not
𝐹𝐹(𝑥𝑥|𝛼𝛼, 𝛽𝛽, 𝜇𝜇) = 10−4 . Equation 1 is then inverted to solve updating the shape parameter when the threshold is reduced
for the stress associated with the SRC POF limit [denoted in the full assessment led to erroneous results. Therefore, the
𝑆𝑆𝑔𝑔 (10−4 )], as shown in Equation 2. Equation 2 corresponds shape parameter was updated in the full assessment in this
to Eq. 21 in HHA-II-3300 [3], with different notation. The study when the threshold was reduced, though this is not
95% one-sided confidence interval lower bounds are currently specified in the Code.
obtained for the shape and scale parameters and used in Points that did not meet grouping criteria at the end of
place of 𝛽𝛽 and 𝛼𝛼 to add conservatism to the distribution prior the equivalent stress list were discarded under the
to the calculation of 𝑆𝑆𝑔𝑔 (10−4 ). assumptions that: (1) these points have the lowest stresses
and will not contribute notably to the POF; and (2) the
1 grouping criterion need to be met for the groups to be valid.
𝑆𝑆𝑔𝑔 (10−4 ) = 𝛽𝛽(− ln(1 − 10−4 ))𝛼𝛼 (2) The Code does not discuss how to deal with this.
𝑆𝑆𝑔𝑔 (10−4 ) is multiplied by the ratio of the flexural mean The analyses performed were done to represent the
2019 and 2021 Code versions as written, conditional on the
strength to the tensile mean strength (𝑅𝑅𝑡𝑡𝑡𝑡 ) to arrive at the
assumptions presented here.
allowable stress value.
The equivalent stress in the component can be 2.3 Finite Element Modeling
computed using the equation outlined in HHA-3213. This 2.3.1 Methodology
formulation is derived from the maximum strain energy For both the simplified and full assessments, FE models
failure theory. The peak equivalent stress value is computed were created and analyzed to obtain the required equivalent
in the component FE model, then compared to the allowable tensile stress outputs. A tensile “dog-bone” geometry was
stress value. used in this mesh refinement study. The finite element
modeling tool, Grizzly [1], was developed at Idaho National
Two criteria must be met to satisfy the simplified Laboratory (INL) and was used to compute the necessary
assessment: stress output for the component. Grizzly is based on the
- 𝐶𝐶𝑚𝑚 < 𝑆𝑆𝑔𝑔 (10−4 ) open-source finite element modeling framework, MOOSE
- 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 < 𝑅𝑅𝑡𝑡𝑡𝑡 ∗ 𝑆𝑆𝑔𝑔 (10−4 ) [2], also developed at INL. Efforts are currently underway
to implement the full and simplified assessment methods
If both criteria are met, the component is accepted per into the Grizzly code as a post-processor.
the simplified assessment.
2.1.3 Full Assessment 2.3.2 Assumptions
The full assessment is primarily outlined in HHA-3217 The FE geometry was selected to match the
and HHA-II-3200. It summarizes the experimental tensile experimental testing done in the Baseline Unirradiated
strength data using the 3-parameter Weibull distribution Graphite Program. In order to reduce compute times, the
given in Equation 1. A shortened version of the steps are as model was simplified to a half symmetry bar, cut on the
follows: plane intersecting the center perpendicular to the axis of
- The threshold parameter is reduced if the component rotation. The grooves on each end, as shown in the drawing,
stress distribution and Weibull strength distribution are used in the mechanical test to grip the specimen and
have too little overlap apply the axial load. In the model, the loading was applied
- The FE integration points are grouped based on: using a pressure boundary condition, and the material at and
o A minimum link (group) volume requirement, above those notches was excluded. A red box is shown in
the drawing to illustrate the portion of the specimen used in
𝑉𝑉𝑚𝑚
the model. Figure 1 below shows the drawings for the tensile
o A minimum relative (group) stress range, Δ
bar and the resulting FE model.
- A probability of survival (1 – probability of failure) is
computed for each group
- The product of the group survival probability (POS) is
calculated
- The POF is found as 1-POS and compared to 10−4
- If POF<10−4 , the component is accepted
HHA-3217 [3] and [5] give more detailed explanations,
which are needed for accurate implementation.

3 © 2023 by ASME
 of geometries, they should be most accurate for the dog-
bone geometry.

In order to consider mesh effects, multiple element

sizes were considered. Five total meshes were created in
order to evaluate the mesh refinement effects on component
acceptance. The coarsest mesh was created with an average
element length of 0.2 inches. That value was cut in half for
each of the subsequent meshes created, such that the final
mesh contained elements with an average length of 0.0125
inches. Figure 2 below shows the five meshes used in the
analysis.

Figure 1: DRAWING FOR TENSILE “DOG-BONE” SPECIMEN. THE

RED BOX OUTLINES THE SECTION OF THE COMPONENT USED
IN THE FE MODEL, WHICH IS SHOWN TO THE RIGHT. DRAWING
DIMENSIONS ARE SHOWN IN INCHES, ALTHOUGH ALL
ANALYSES USED METRIC UNITS.

The diameters at the large and small ends are 19.3 and Figure 2: FIVE MESHES USED IN MESH CONVERGENCE STUDY
8.76 mm, respectively. The length of the model shown is FOR THE DOG-BONE TENSILE SPECIMEN.
41.02 mm, the length of the small diameter portion is 22.225
mm and the radius of curvature on the fillet is 25.4 mm. In order to evaluate the convergence of stress with mesh
Two nodes on the large surface were constrained to fix refinement in the component, the equivalent stress was
translation and rotational modes. Zero axial displacement measured along the center axis of the model using average
was applied at the small surface for symmetry, and a element average values. The equivalent stress formula is
uniform pressure load was applied at the large surface in defined in HHA-3213 of [3]. The stress along the line for
order to axially load the specimen in tension. each model was plotted in Figure 3.

2.4 Mesh Refinement Study

An important aspect of FE modeling that must be
considered is the effect of mesh size on the solution. In
general, smaller elements will provide a more accurate
solution; however, models with a significant number of
elements can quickly become very expensive in terms of
computational costs. The analyst is therefore required to
consider costs vs. benefits when selecting a mesh for his or
her model, where computational requirements are weighed
against model resolution.
In order to address this issue, it is common practice for
the analyst to conduct a mesh convergence study. In such a
study, sequential meshes are created where each model
contains smaller elements than the last. All of the meshes
are then analyzed with identical parameters and conditions.
The solutions from the various meshes are compared to Figure 3: PLOT OF EQUIVALENT STRESSES USING ELEMENT
identify the element size at which further refinement has AVERAGE VALUES THROUGH THE CENTER OF THE MODEL.
little effect on the results. It is then up to the modeler to
select the appropriate element size for use in the analysis With a load of 1 MPa applied to the top surface, the
using the mesh convergence study to justify that decision. analytical solution in the gauge region is 4.853 MPa. All
The dog-bone tensile component was selected for this meshes resulted in a stress at the end surface of 4.853 MPa,
analysis because the Weibull strength distribution used in but the increased mesh refinement models provided higher
the full and simplified assessments is a tensile strength
distribution. While the assessments are used with a variety

4 © 2023 by ASME
 resolution in the portion of the model where the cross-
sectional area changes.

2.5 Data
2.5.1 Experimental Strength Data
Experimental strength data were used from the
Unirradiated Graphite Testing Baseline Program for two
graphite grades of IG-110 and PCEA tensile test dog-bone
specimens. A total of 290 IG-110 specimens from two
billets and a total of 127 PCEA specimens from a single
billet (billet 01D336, referred to as PCEA36), were used to
determine the corresponding Weibull strength distributions.
Table 1 displays the Weibull lower bound distributions used
in this analysis for each grade.
Figure 5: PCEA36 DISTRIBUTIONS
Table 1: WEIBULL PARAMETER ESTIMATES
2.5.2 FE Model Component Stress Data
Assessment 𝜇𝜇 𝛼𝛼 𝛽𝛽 The finite element model for the tensile dog-bone specimen
Grade
was created under the assumption of purely elastic behavior.
Lower Lower
Therefore, once a stress state has been determined in the analysis,
Bound Bound
the magnitudes for each stress can be scaled by the load of
IG-110 Simplified 0 14.804 27.515 interest.
Full 15.129 6.2645 12.292 The values of tensile strength collected during the
PCEA36 Simplified 0 5.7433 19.273 mechanical testing of these components correspond to the
Full 9.9526 2.4870 8.947 maximum engineering stress experienced in the small diameter
section of the sample at the moment immediately prior to failure.
Figures 4 and 5 have many aspects pertaining to the The component stresses in the model can be scaled to match
distributions. The dark grey columns represent histograms whatever strength value is of interest in the study.
of the underlying experimental data distributions. The green For this analysis, fully integrated 8-noded elements were
colors indicate full assessment distributions and the blue used. The integration point values for each element in the mesh
colors indicate simplified assessment distributions. The were combined to a weighted element average for each principal
light colors are the best fits to the data, while the dark colors stress value. For each element, the three principal stresses were
represent the conservative 95% lower bound distributions combined in accordance with HHA-3213 to form an average
prescribed for computation in the Code. The dark blue solid equivalent stress for each element. It is worth noting that this
vertical line is 𝑆𝑆𝑔𝑔 and the dashed dark blue vertical line is may not be the process that is directly outlined in the ASME
𝑅𝑅𝑡𝑡𝑡𝑡 ∗ 𝑆𝑆𝑔𝑔 . Note that 𝑆𝑆𝑔𝑔 is found using the lower bound Code, which states that values are to be determined at each
distribution for the simplified assessment. The solid dark integration point, and the associated volume for that point is to
green line is the stress at a 𝑃𝑃𝑃𝑃𝑃𝑃 = 10−4 using the full be used in the grouping step.
assessment lower bound distribution, for reference. Although the Code specifies that this procedure be done on
integration points, it is common practice to use element volumes
and element-averaged quantities. Development work is currently
underway in Grizzly to implement the simplified and full
assessments so that the FE model can directly output peak
equivalent stress and POF values.

2.6 Dog-bone Analysis

Dog-bone geometries and graphite grades IG-110 and
PCEA36 were used for comparison. The results are compared for
FE applied to high and low loads in the full and simplified
assessments. The low loads were found to straddle the 10−4
limit.
𝑉𝑉𝑚𝑚 and Δ criteria were changed between the 2019 and 2021
Code versions in the full assessment; therefore, both were
Figure 4: IG-110 DISTRIBUTIONS considered. Table 2 provides the 𝑉𝑉𝑚𝑚 and Δ criteria for both the
2019 and 2021 Code versions, for both grades. 𝑉𝑉𝑚𝑚 was calculated

5 © 2023 by ASME
using 𝐾𝐾1𝑐𝑐 (
𝑀𝑀𝑃𝑃𝑎𝑎
) values of 1.33 for PCEA36 and 1.07 for IG- Table 5: SIMPLIFIED ASSESSMENT ACCEPTANCE METRICS FOR
√𝑚𝑚 IG-110 AND PCEA36
110.
Grade/Metric 𝑆𝑆𝑔𝑔 (10−4 ) MPa 𝑅𝑅𝑡𝑡𝑡𝑡 ∗ 𝑆𝑆𝑔𝑔 (10−4 ) MPa
Table 2: VOLUME GROUPING CRITERIA FOR IG-110 AND PCEA
IG-110 14.77 19.386
Grade/ 2019; Δ = 0.07 2021; Δ = 0.2 PCEA36 3.877 6.188
Code Version
Peak equivalent stresses are listed below in Tables 6 and 7
IG-110 𝑉𝑉𝑚𝑚 (𝑚𝑚𝑚𝑚3 ) 0.008 14.704
from each of the five dog-bone meshes for IG-110 and PCEA36
PCEA 𝑉𝑉𝑚𝑚 (𝑚𝑚𝑚𝑚3 ) 512 572.104
at the two target membrane stresses applied in this study. The
high applied loads result in failure at the membrane stress check.
The scope of this study contains comparisons according to The low applied load was set to the 𝑆𝑆𝑔𝑔 limit, allowing passing of
Table 3.
the membrane stress check. However, components failed the
peak equivalent stress check, regardless of mesh refinement.
Table 3: STUDY MATRIX
For IG-110, the relative change in peak equivalent stresses
Target POF IG-110 PCEA36 between each level of mesh refinement was <2%. Both applied
High 2019 2021 2019 2021 stresses resulted in peak equivalent stresses above the 𝑅𝑅𝑡𝑡𝑡𝑡 ∗ 𝑆𝑆𝑔𝑔 =
Low 2019 2021 2019 2021 6.188 𝑀𝑀𝑀𝑀𝑀𝑀 limit, therefore mesh refinement did not affect
acceptance for these applied stresses. However, with peak
The fitted quantiles used to scale the FE applied load for equivalent stresses increasing by nearly 1 MPa between the least
PCEA36 and IG-110 are summarized in Table 4. For a realistic and most refined meshes, one could conceivably apply a stress
application of the Code, different applied pressure loadings in which mesh refinement could affect the results of the second
would not be applied for the full and simplified assessments for check in the simplified assessment.
a given component. Different applied pressure loadings were
Table 6: PEAK EQUIVALENT STRESSES FOR IG-110
used for illustration purposes in the full and simplified
assessments in this paper. As expected, target applied pressure
Membrane 1 2 3 4 5
loadings associated with the higher quantile are very similar
Stress/Mesh
when using the 2- and the 3-parameter Weibull distributions
14.77 MPa 14.812 14.866 15.194 15.411 15.493
(simplified and full). However, the non-zero threshold parameter
greatly shifts the target pressure loading to the right in the left
tail of the distribution (near the target POF=10−4 ). 18.082 MPa 18.133 18.199 18.601 18.867 18.968
Table 4: GRAPHITE GRADES AND APPLIED PRESSURE
LOADINGS (MEMBRANE STRESS)
The conclusions made in Table 6 for IG-110 are also
Low Target Simplified (MPa) Full (MPa) applicable to the results for PCEA36 in Table 7.
IG-110 14.77 15.5
PCEA36 3.877 6.2 Table 7: PEAK EQUIVALENT STRESSES FOR PCEA36
High Target
Membrane 1 2 3 4 5
IG-110 26.842 26.723
Stress/Mesh
PCEA36 18.082 17.6734
3.877 MPa 3.888 3.901 3.986 4.042 4.063
18.082 MPa 18.134 18.196 18.589 18.851 18.950
3. RESULTS AND DISCUSSION
3.1 Results 3.1.2 Full Assessment
3.1.1 Simplified Assessment The full assessment is performed by comparing a
The dog-bone meshes showed slightly increasing peak component POF to the SRC POF limit. In general, the change in
equivalent stresses with increased mesh refinement because of POF is minimal for the finest two meshes. This POF trend did
elevated stresses at the fillet radius. Whether the changes in mesh not entirely hold for IG-110 when applying the 2019 Code rules.
refinement affect the acceptance decision is dependent on how It is hypothesized that the small 𝑉𝑉𝑚𝑚 is still causing enough
close the applied stress in the FE model is to the allowable stress changes in the POF to warrant further mesh refinement before
value. Table 5 provides 𝑆𝑆𝑔𝑔 (10−4 ) and the allowable stress value saying the POF has converged. Figure 6 displays the lower
[𝑅𝑅𝑡𝑡𝑡𝑡 ∗ 𝑆𝑆𝑔𝑔 (10−4 )] for both grades in MPa. pressure loading results. The black dashed lined is the

6 © 2023 by ASME
POF=10−4 limit. Results at the low pressure loadings show how
mesh refinement can affect the acceptance decision.

Figure 6: FULL ASSESSMENT RESULTS FOR THE EFFECTIVE

0.0001 APPLIED LOADINGS

Figure 7 displays the results at the higher pressure loadings. The

loads were high enough to not affect the acceptance decision.
However, the plot displays that POF convergence did not happen
until the 3rd or 4th level of mesh refinement.

Figure 8: FRACTIONAL CHANGE IN POF AS A FUNCTION OF

MESH

3.2 Discussion
This mesh convergence study showed slightly increasing
peak stress values around the surface of the fillet region of the
component with increased mesh refinement. This behavior is
common in models where there are abrupt changes in the
geometry, such as sharp corners, holes, fillets, and notches. The
dog-bone geometry considered for analysis in this paper contains
Figure 7: FULL ASSESSMENT RESULTS FOR THE EFFECTIVE 0.5
a fillet where the large cross-section is reduced to the smaller
APPLIED LOADINGS cross-section found in the center of the sample. This is
commonly referred to as a stress riser or stress concentration,
Figure 8 displays the fractional change in POF as a function where the computed stresses in the FE model are elevated due to
of the fractional change in peak equivalent stress. Fractional geometric factors. This can be problematic when considering the
change for arbitrary x relative to previous x0 is: |
𝑥𝑥−𝑥𝑥0
|. The simplified assessment as only the peak equivalent stress is
𝑥𝑥0 considered in the second check. In practice, only the neck portion
change in POF converges to less than one between meshes 4 and of the dog-bone is important when applying the tensile test. The
5 for both pressure loadings. methods applied in this study used the entire dog-bone FE model,
rather than sub-setting to the neck portion in order to illustrate
how complex geometries may affect the mesh refinement
necessary when applying the simplified and full assessments.
Stress concentrators may require more mesh refinement
than is typical to get convergence in simplified assessment
results. In such cases, where the peak stress is near the allowable
stress value, it would be wise to consider using the full
assessment for further investigation. However, if the peak

7 © 2023 by ASME
stresses are far lower than the acceptable limits, it may not be the 10−4 quantile of the distribution relative to the
necessary to further refine the mesh. original 3-parameter Weibull distribution. This is
One may think the grouping of elements in the full believed to work as intended in the full assessment. In
assessment would mask any mesh refinement effects. Though addition to reasons for the threshold reduction
not conclusive, this study suggests that the magnitude of 𝑉𝑉𝑚𝑚 identified in the first bullet, the threshold is reduced to
relative to the mesh element volumes influences the degree to account for the fact that enough data are not collected
which mesh refinement affects the POF. 𝑉𝑉𝑚𝑚 is substantially to characterize the left tail of the Weibull distribution
smaller in the 2019 Code version for IG-110, and we see the POF well experimentally, and to add in conservatism since
continuing to rise with mesh refinement for the lower applied the Code does not prescribe using a lower bound on the
loadings (10−4 ), whereas the POF appears to level off for threshold parameter.
Meshes 3-5 for all other cases. This suggests that the larger 𝑉𝑉𝑚𝑚 - The “new” or reduced 3-parameter Weibull distribution
values result in POFs that are less sensitive to the micro-changes changes slightly depending on the peak equivalent
in element stresses that come with continued mesh refinement. stress. Since mesh refinement affects peak equivalent
This effect only appeared with the lower pressure loading. stress, slightly different reduced Weibull distributions
As mentioned, the current HHA-3000 Code sections do not were used in the final POF calculation depending on
provide guidelines for mesh refinement. However, that is not mesh refinement. Though, when looking into this
uncommon. For example, Subpart B on Composite Materials [3] further, differences in reduced 3-parameter Weibull
utilizes a similar methodology as the simplified assessment for distributions were minimal in terms of parameters used.
component acceptance and does not provide guidelines for mesh - Finally, Figures 4 & 5 illustrate that the relationship
refinement. Depending on the geometry and the loading applied, between the simplified assessment limits and the
stress convergence within a model may not equate to POF stresses associated with the POF=10−4 in the 3-
convergence within that same model. Additional mesh parameter Weibull lower bound distribution in the full
refinement studies are currently underway that evaluate these assessment can change by graphite grade. These are not
statements for other geometries and graphite grades. artifacts of the methodologies used in the assessments,
Though this paper focused on the mesh refinement effects but rather in the material properties as they present
on acceptance decisions in the full and simplified assessments, themselves in the Weibull fits to the experimental data.
other items of interest pertinent to the methodologies are
illustrated in this study. Those include: 4. CONCLUSION
This study set to determine mesh refinement effects on the
- When searching for a load that resulted in a POF=10−4 acceptance decision of nuclear grade graphite core components
in the full assessment for PCEA36, initially the 10−4 per Article HHA-3000 of the ASME BPV Codes [3]. Five mesh
load based on the 3-parameter lower bound distribution refinement levels were considered, with the peak equivalent
was used, around 10 MPa. This resulted in a POF stress changing by less than 2% for the three finest meshes.
~0.007 for Mesh 3. However, in an experiment where Mesh refinement did not affect the simplified assessment
the threshold was not reduced, the peak equivalent acceptance decisions for the applied pressure loadings in this
stress quickly became less than the original threshold study. However, since the peak equivalent stresses changed by
parameter as the load of 10 MPa was reduced, in an nearly 1 MPa between the least and most refined meshes, it is
attempt to achieve a load associated with a POF=10−4 . conceivable that a case exists where mesh refinement could
The peak equivalent stress less than the threshold affect the acceptance decision in the simplified assessment.
results in a POF=0 when the threshold reduction step is Mesh refinement affected the full assessment acceptance
removed. It is hypothesized that one reason for the decisions for the low applied pressure loadings in this study for
threshold reduction step is to avoid situations like this, both grades, even though peak equivalent stresses changed by
which absent the threshold reduction step would not less than 2%. POF did not hit convergence until the 3rd or
have allowed the component to obtain a POF between sometimes 4th mesh refinement. This suggests that FE modelers
0 and ~0.007. need to evaluate POF convergence as a criteria for sufficient
- In the full assessment, originally, applied stresses at the mesh convergence running components through the full
10−4 target POF were found using the 3-parameter assessment. It is hypothesized that smaller minimum link
Weibull lower bound distribution. However, the volumes necessitate more extensive mesh refinements to
resulting POF’s were much larger than 10−4 (by at least achieve POF convergence due to sensitivities in micro-changes
one order of magnitude). The major reason for this is to mesh element stresses that appear during mesh refinement.
that the 3-parameter Weibull lower bound distribution This study suggests POF convergence appears to happen after
is used at the beginning of the full assessment, however, stress distribution convergence, contrary to previous work done
after the threshold reduction step, a new 3-parameter by Hindley in [4]. POF convergence needs to be considered
Weibull distribution is used in the final POF when determining adequate FE stress distribution convergence,
calculations. Since this new distribution has a decreased especially when the calculated component POF is near the SRC
threshold, that can greatly affect the amount area below limit.

8 © 2023 by ASME
ACKNOWLEDGEMENTS
The authors would like to thank those that contributed to the
Unirradiated Graphite Testing Baseline Program, for providing
the necessary data, as well as the Design Task Group members
for the shared knowledge and understanding of these
assessments.

REFERENCES
[1] Benjamin W. Spencer, William M. Hoffman, Sudipta
Biswas, Wen Jiang, Alain Giorla, and Marie A. Backman,
(2021). “Grizzly and BlackBear: Structural component aging
simulation codes,” Nuclear Technology, 207:7, 981-1003, DOI:
10.1080/00295450.2020.1868278
[2] Cody J. Permann, Derek R. Gaston, David Andrš, Robert
W. Carlsen, Fande Kong, Alexander D. Lindsay, Jason M.
Miller, John W. Peterson, Andrew E. Slaughter, Roy H. Stogner,
Richard C. Martineau, (2020). “MOOSE: Enabling massively
parallel multiphysics simulation,” SoftwareX, Volume 11,
100430, ISSN 2352-7110.
[3] American Society of Mechanical Engineers (ASME)
Boiler and Pressure Vessel Code (BPVC), Section III, Rules for
the Construction of Nuclear Facility Components, Division 5,
High Temperature Reactors, (2011). New York.
[4] Michael P. Hindley, Mark N. Mitchell, Christiaan
Erasmus, Ross McMurtry, Thorsten H. Becker, Deborah C.
Blaine, and Albert A. Groenwold, (2013). “A numerical stress
based approach for predicting failure in NBG-18 nuclear
graphite components with verification problems,” Journal of
Nuclear Materials, 436(1-3): pp. 175-184.
[Link]
[5] Josina W. Geringer, Timothy D. Burchell, Mark
Mitchell, (2021). “ASME NTB: Background information for
addressing adequacy or optimization of ASME BPVC Section
III, Division 5 rules for nonmetallic core components.”
[6] T. R. Allen, K. Sridharan, L. Tan, W. E. Windes, J. I.
Cole, D. C. Crawford, Gary S. Was, (June 2008). “Materials
challenges for Generation IV nuclear energy systems,” Nuclear
Technology, Vol. 162(3), pp. 342-357.
[7] B.J. Marsden, A.N. Jones, G.N. Hall, M. Treifi, P.M.
Mummery, “Graphite as a core material for generation IV
nuclear reactors,” in: Structural Materials for Generation IV
Nuclear Reactors, Elsevier, 2017, pp. 495–532,
[Link]
[8] J. Kane, C. Karthik, D. P. Butt, W. Windes, and R. Ubic,
(2011). “Microstructural characterization and pore structure
analysis of nuclear graphite,” Journal of Nuclear Materials, 415
(2) 189-197. [Link]
[9] C. Karthik, J. Kane, D. P. Butt, W. Windes, and R. Ubic,
(2012). “Microstructural characterization of next-generation
nuclear graphites.” Microscopy and Microanalysis, 18(2) pp.
272-278.
[10] D. Kanse, I.A. Khan, V. Bhasin, R.K. Singh, (2015).
“Interpretation of ASME Code rules for assessment of graphite
components”. SMiRT23, Division II, Paper ID 346.

9 © 2023 by ASME