canada’s nuclear science and technology

canada’s nuclear science and technology
Issue 110 / Summer 2012
Volume 1, Number 1
June 2012
canada’s
nuclear
science and
technology
journal
www.aecl.ca
AECL Nuclear Review
ABOUT AECL
AECL Nuclear Review showcases innovative and
important nuclear science and technology that is aligned
with AECL’s core programs. The Journal welcomes
original articles and technical notes in a variety of subject
areas: CANDU Nuclear Industry; Nuclear Safeguards and
Security; Clean Safe Energy including Gen IV, Hydrogen
Technology, Small Reactors, Fusion, Sustainable Energy
and Advanced Materials; Health, Isotopes and Radiation;
and Environmental Sciences. The accepted peer-reviewed
articles are expected to span different disciplines such as
engineering, chemistry, physics, and biology.
Atomic Energy of Canada Limited (AECL) is Canada’s leading
nuclear science and technology laboratory. For over 60
years, AECL has been a world leader in developing peaceful
and innovative applications of nuclear technology through
its expertise in physics, metallurgy, chemistry, biology and
engineering.
acting editor
editorial board
AECL Nuclear Review welcomes Canadian and international
research scholars and scientists from different disciplines
to its new publication, which reflects the integration of
scientific researchers and industrial practitioners.
G.L. Strati, Manager, Mechanical Equipment
Development, AECL
associate editors
F.M. Courtel, Chemistry, AECL
A. Khalifa, Mechanical Engineering, AECL
Today, AECL continues its commitment to ensure
that Canadians and the world receive energy, health,
environmental and economic benefits from nuclear science
and technology with confidence that nuclear safety and
security are assured.
C. Butler, Manager, R&D Operations, AECL
R. Didsbury, General Manager, R&D Operations, AECL
G.L. Strati, Manager, Mechanical Equpment
Development, AECL
B. Sur, Director, Nuclear Science, AECL
Copyright © 2012 Atomic Energy of Canada Ltd. All rights
reserved.
Volume 1, Number 1
June 2012
invited articles
Science and Technology Enabling Canada’s Tier 1 Nuclear Sector
R. Walker
-1-
Assessing Long-Term Performance of CANDU® Out-of-Core Materials
R.L. Tapping, Y.C. Lu, D. Mancey and Z.H. Walker
-5-
full articles
Localized Corrosion of Nuclear Grade Alloy 800 Under Steam
Generator Layup, Startup and Operating Conditions
Y. Lu
- 13 Nuclear Data and the Effect of Gadolinium in the Moderator
J.C. Chow, F.P. Adams, D. Roubstov, R.D. Singh, and M.B. Zeller
- 21 -
Measurements of the High Dose Rate Profiles Inside a Shutdown CANDU® Reactor
C. Jewett, J. Chow, D. Comeau, G. Jonkmans, B. Smith, B. Sur, D. Taylor, S. Yue
- 27 Optimization of the Spatial Mesh for Numerical Solution of the
Neutron Transport Equation in a Cluster-Type Lattice Cell
R.S. Davis
- 35 -
Self-Excited Acoustic Resonance of Isolated Cylinders in Cross-Flow
A. Mohany
- 45 -
technical notes
A Novel Boron-Loaded Liquid Scintillator for Neutron Detection
G. Bentoumi, X. Dai, E. Pruszkowski, L. Li, B. Sur
- 57 Inverse Collimator-Based Radiation Imaging Detector System
A. Das, B. Sur, S. Yue, G. Jonkmans
- 61 -
Anthropogenic Radionuclides in Ottawa River Sediment near Chalk River Laboratories
D.J. Rowan
- 67 -
science and technology enabling canada’s tier 1
nuclear sector
Dr. Robert Walker, President & CEO, AECL
Atomic Energy of Canada Limited, Chalk River Laboratories, Chalk River, Ontario, Canada, K0J 1J0
Welcome to the inaugural issue of Atomic Energy of Canada
Limited’s AECL Nuclear Review. We at AECL are dedicating
this journal to the dissemination of the latest science and
technology (S&T) information through the work of AECL
and its partners, and by so doing, we will contribute to the
advancement of the peaceful, safe and trusted use of nuclear technology for the betterment of Canada and the world.
AECL is Canada’s premier nuclear science and technology organization. 2012 is a milestone year for AECL as it
celebrates 60 years as a world leader in the development
and deployment of peaceful and innovative applications of
nuclear technology, enabled by its deep expertise in physics, metallurgy, chemistry, biology, ecology and engineering.
Through its journey, AECL has contributed to the creation
and sustainment of a strong domestic nuclear sector. Our
nuclear sector today, without a doubt, has a strong foundation on which to build.
However, as we in the sector reflect on its future, we see
immediately that today’s world, which sets today’s context
for civil nuclear technology and nuclear energy, is far different than it was for much of the sector’s previous history.
Globalization, interconnected societies, climate change –
these are each terms used to characterize today’s reality.
Our world has become smaller, interdependent, and more
complex, with an accelerating pace of change. Science and
technology, or more explicitly the globalization of science
and technology, has contributed directly to this new reality.
The resultant dynamics in societies and nations are challenging leaders to adapt public policy to address simultaneously the competing priorities of public-sector debt and
deficits, sustainable economic growth and the evolving
needs of public health, safety and security. But with these
challenges also comes opportunity – opportunity for those
sectors that recognize and leverage their strategic advantage in a manner appropriately adapted to today’s realities.
Canada is a Tier 1 Nuclear Nation
To chart a path to the future, one must first understand
where one is today.
Canada is a Tier 1 Nuclear Nation - one of a small number of
countries with a comprehensive nuclear sector. Canada is
committed to nuclear energy, with nuclear utilities operating in three provinces. Our utilities have been consistently
ranked as among the best in the world for safe, reliable performance. The standards are rightly very high. Our nuclear
utilities reside in host communities where mutual respect
and trust have been established and sustained. Canada has
an internationally respected, independent regulator operating in a coherent legislative and legal framework that
AECL NUCLEAR REVIEW
1
aecl Nuclear Review
vol 1, Number 1, june 2012
science and technology enabling canada’s tier 1 nuclear sector
dr. robert walker
meets and exceeds international expectations. Canada is a
sought-out and active contributor to the advancement and
implementation of international nuclear safety and security
policies and standards. Canada has a robust domestic supply chain from mining through manufacturing and services,
anchored by a domestic nuclear reactor original equipment
manufacturer (OEM). Canada has a vibrant network of colleges and universities that are educating our next generation of nuclear workers while providing the appropriate research to augment their training. And Canada has the AECL
Nuclear Laboratories, the home of critical nuclear science
and technology expertise and licensed infrastructure that
contributes to Canada’s nuclear policy, regulatory framework, operational effectiveness, education and technological capabilities.
Attributes of a Tier 1 Nuclear Nation that Build Strategic
Advantage
A Tier 1 Nuclear Nation today; but what of tomorrow?
An answer to this question emerges through an assessment
of the attributes – what “good” looks like - that would provide or sustain strategic advantage for our nuclear sector
into the future. Following are a possible set of such attributes around which the nuclear sector could rally its collective efforts.
A Tier 1 Nuclear Nation has a supply chain that is internationally competitive, with a demanding domestic market to
help establish its credentials. Being internationally competitive requires that we understand what provides this
competitive advantage. It also requires an understanding
of the unique social, economic and public policy needs of
the market. Our domestic nuclear platform, the CANDU®, is
particularly noteworthy; CANDU has a strategic advantage.
It is important that we understand and leverage that strategic advantage in domestic and international markets, recognize how it is complementary to other nuclear platforms,
and appreciate how it plays into the public-policy priorities
of customer nations.
A Tier 1 Nuclear Nation has a credible regulator who attests
to the world that the country’s nuclear solutions are safe. A
necessary ingredient to sustained public trust is a credible
regulatory system founded in fact; good science is a core ingredient to good regulation. There are many issues where
the voice of science is needed to both inform regulatory
decisions and to underpin the technological solutions that
must respond to regulatory direction.
2
A Tier 1 Nuclear Nation has a sustained supply of highly
qualified people. This supply must be enabled by appropriate education, economic development and immigration policies and systems. We aspire to attract Canada’s
best and brightest into careers that span public policy and
regulation, operations, education, engineering and manufacturing and research, science and technology. Sustaining highly qualified people is a matter of both quantity and
quality; they must be supported by a system of enablers
that promote the movement of people across government,
academic and industry sectors.
A Tier 1 Nuclear Nation values innovation. Focussing our sector on technological innovation is enormously important,
particularly in such a capital-intensive industry as ours, but
indeed innovation has many aspects to it. Business innovation is about how we work and how we translate good ideas
into operational or commercial success. For example, the
concept of a performance-driven supply chain is a business
innovation. As opposed to a customer saying to its suppliers “Here is a technical specification, please build to this
specification”, our industry could do as the aerospace and
automotive sectors did decades ago, that is, say “Here is the
performance objective, please provide the best solution to
meet this performance within our stringent safety and security requirements”. This approach rewards innovation
by setting conditions that stimulate increased productivity
and profitability. It also affords the opportunity to leverage
our sector’s unique technological advances for competitive
advantage in other nuclear platform supply chains or in
other markets.
A Tier 1 Nuclear Nation is resilient in the face of volatility
in the market. The reality is that the nuclear-energy business tends to fluctuate episodically with new build and
refurbishments as intensive a periodic opportunities interspersed with services. The challenge is how to sustain sector capabilities and capacity to be responsive and resilient
to that reality. For example, a supply chain that is able to
lever its technology, production and service offerings into
non-nuclear sectors will expand its market space and sustain growth while building competitive advantage.
A Tier 1 Nuclear Nation exploits public-private partnerships.
The nuclear business is capital intensive, with long lead
times for project design through implementation. Access
to long-term “patient” capital is a necessary ingredient to
success. Innovative business models that bring together
the strengths of the public and private sectors in managing
program risk can provide competitive advantage.
A Tier 1 Nuclear Nation builds and sustains public trust. Public trust takes time to build, and can be all too easily lost. A
simple reality must be acknowledged - humans have a natural inclination to fear radiation. Trust must be built on the
principles of transparency, accountability, effective regulation, objective discourse – and on accessible science.
A Tier 1 Nuclear Nation has reliable domestic access to nuclear science and technology expertise and infrastructure.
The advancement of nuclear technology, the sustainment of
aecl Nuclear Review
vol 1, Number 1, june 2012
science and technology enabling canada’s tier 1 nuclear sector
dr. robert walker
nuclear operations and the regulation of both the technology and operations require large-scale licensed science and
technology facilities. These facilities must be capitalized
and sustained through an appropriate sharing of publicand private-sector investments.
An assessment of Canada’s status today relative to these attributes identifies both strengths and opportunities for improvement. How do we nourish our strengths and address
the areas for improvement? One approach is through the
effective use of science and technology.
Science and Technology Enables Strategic Advantage
Science and technology, appropriately harnessed, can be a
key contributor to the realization of the above attributes,
thereby enabling the strategic advantage of our nuclear
sector. But where and how do we align effort across the
science and technology community - and the investments
being made in the efforts of this community - to help ensure
we are sustaining our Tier 1 status?
Here are some thoughts regarding a framework for helping
build this alignment, with reference to specific examples of
how AECL’s science and technology efforts are contributing.
S&T enables sector competitivity through CANDU’s strategic advantage. CANDU’s strategic advantage is rooted in
the flexibility of its fuel cycles, which in turn provides the
opportunity for CANDU to complement, not compete with,
alternative nuclear reactor designs in markets where this
fuel cycle flexibility is of strategic value. For example, AECL
has continually strived to evolve the current CANDU reactor design, particularly through a program called CANDU-X,
which evolved the CANDU design over the years into a supercritical water-cooled reactor concept, which meets the
objectives of the Generation IV International Forum; a safer,
sustainable, economical and more proliferation resistant
and physically secure design that retains CANDU’s original
fuel cycle advantage.
S&T enables public trust through science-informed regulation. In light of the March 2011 Fukushima Diachi reactor accident, there is renewed regulatory focus on nuclear
emergency response and on the long-term effects of low
level radiation on humans and non-human biota. For example, AECL is working with a broad set of partners to advance
the scientific community’s understanding of the biological
mechanisms associated with the long-term exposure of
biota to low-level radiation in the environment. This scientific knowledge is critical to establishing the factual basis
for regulation and for managing the response to nuclear incidents in the future.
S&T enables sector capability through its experiential
learning for its highly qualified workforce. Engagement in
nuclear science and technology activities is a powerful
mechanism of experiential learning for students, policy
makers, regulatory staff, operators, industrialists, scientists,
engineers and technologists. For example, AECL’s approach
to the execution of its comprehensive suite of S&T activities
is to value collaboration across a broad set of nuclear sector partners, as well as to make AECL’s unique nuclear S&T
facilities accessible to others for their needs.
S&T enables business innovation through a performancedriven supply chain. Science and technology helps translate nuclear utilities’ and OEM performance specifications
into cost-effective technological solutions, demonstrating
that these solutions meet demanding safety requirements
while at the same time driving commercial competitive advantage. For example, AECL is expanding its efforts to help
Canadian Small- and Medium-Sized Enterprises (SMEs) in
the nuclear supply chain not only advance their technological solutions but also validate that these solutions meet and
exceed regulatory requirements.
S&T enables sector resilience through the development of
dual-use technologies. Technological solutions developed to
meet the very demanding requirements of the nuclear environment can also provide competitive advantage in other
industry sectors. For example, AECL has a rich history of
working with suppliers in the development of robotic tooling for nuclear reactor maintenance and repair. These robotic solutions have potential application in the harsh environments found in other industries, such as petro-chemical
and mining.
S&T enables a holistic perspective through an integrated lifecycle outlook. The cost effectiveness of nuclear technology
in general, and of nuclear energy in particular, requires a
holistic view of the entire lifecycle including mining and fuel
production, plant licensing, new build, operation, refurbishment, decommissioning and waste management. Science
and technology can help inform the entire lifecycle design
to improve overall system economics and to help identify
the substantial cost and effectiveness drivers. AECL has the
integrated S&T expertise and capabilities to nurture this
holistic view within Canada’s nuclear sector.
Conclusion
Canada has a rich history of success in nuclear technology,
built over many decades. It is a solid foundation on which
we can build. Sustaining Canada’s Tier 1 nuclear status
will require a shared vision of where we want to take our
nuclear sector into the future. Science and technology has
a critical role to play in this journey and AECL, through this
new journal, intends to share information on the latest S&T
advances that are helping our sector along this journey.
AECL NUCLEAR REVIEW
3
4
INVITED ARTICLE
Abstract
As so-called second-generation power
reactors are approaching the end of their
original design lives, assessments are being
assessing long-term performance
of candu® out-of-core materials
R.L. TappingA*, Y.C. LuA, D.S. ManceyA and Z.H. WalkerB
A
B
Atomic Energy of Canada Limited, Chalk River Laboratories, Chalk River, Ontario, Canada, K0J 1J0
Candu Energy Inc., 2280 Speakman Drive, Mississauga, Ontario, Canada, L5K 1B1
in terms of in-service ageing and continued
Article Info
Keywords: steam generator corrosion, flow accelerated corrosion, carbon steel corrosion, CANDU® materials
Article history: Received 13 April 2012, Accepted 7 May 2012, Available online 30 June 2012.
* Corresponding Author: (613) 584-3311 ext. 43219, [email protected]
fitness for service, ageing of out-of-core
1. Introduction
made to determine the feasibility and
economics of extending plant life. Although
components exposed to neutron and gamma
irradiation are often those of most concern
components can also limit the possibility of
extended service life beyond design life. In
CANDU® reactors, life extension decisions
occur when the Zr-2.5Nb pressure tubes
reach end of life, typically after about 25
years of service for the first CANDU-6 units.
At the time of pressure tube replacements,
the remaining life predictions for several
other major components or systems provide
the information required to determine
life extension feasibility. Several CANDU
reactors are currently being refurbished,
with others planned, and experience to
date shows that the steam generators,
heat transport system piping and various
balance of plant piping systems are typically
those requiring careful assessment to ensure
successful refurbishment. In this paper, we
discuss AECL R&D that is oriented towards
providing the chemistry and materials
inputs required to assess current condition
and predict future ageing of CANDU reactor
out-of-core components and systems, and
in particular steam generators (Alloy 800
tubing and carbon steel internals), feeder
pipes and related heat transport system
piping (carbon steel flow accelerated
corrosion, feeder cracking). Systems and
components that may impact future life will
also be discussed, along with the related
R&D, and this includes balance of plant
system piping (feedwater piping and buried
piping), cables and concrete structures.
A significant challenge in determining the feasibility of extending the life of nuclear power plants is to quantitatively predict remaining life, up to 60 to 80 years
of service, of existing components after 20 to 30 years of service to date. In some
cases, there is no in-service record of degradation, which introduces uncertainty
in predictive approaches that can be used to assess future component and system
performance. Laboratory data are often conservative in predicting materials performance, although there are striking examples where laboratory data predicted
in-service corrosion, for instance, the work of Coriou, which highlighted the susceptibility of Alloy 600 to stress corrosion cracking (SCC) in hot pure water [1].
In other cases, in-service degradation has occurred, which was not predicted by
mechanistic studies. An example of this is the carbon steel cracking in primary heat
transport system (PHTS) water at one CANDU plant, which remains unexplained
mechanistically. Further, much of the laboratory data have been obtained using
essentially static conditions in the sense that the test environment did not change
significantly with time. It is well established that actual plant operation is such
that key components such as SGs and PHTS components are subjected to fluctuating conditions; thus, over long operating lives, these components experience many
changes in pH, chemistry, electrochemical corrosion potential (ECP), mechanical
loading and temperature. The accumulation of these stresses and especially synergistic effects between them, on component condition has not been evaluated
with laboratory testing but remains an issue for long term performance. Some approaches to this for SG tubing will be discussed.
In this paper, the focus will be on two key materials and issues; one is Alloy 800
steam generator (SG) tubing performance in CANDU service, and recent results
from PWR reactor SG experience, which have resulted in a need to re-evaluate the
R&D required to address this operating experience. The other is flow accelerated corrosion (FAC) and intergranular stress corrosion cracking (IGSCC) of carbon
steel in various CANDU systems and components. These topics will be addressed
in some detail, but other materials’ aging and degradation issues of interest for the
longer term will be mentioned, including those related to electrical cables, concrete
structures and balance of plant piping, and the associated R&D in place to assess
these components.
2. Alloy 800 Steam Generator Tubing
After more than 30 years of corrosion-free in-service experience with Alloy 800,
there have been recent reports of secondary side IGSCC detected in this material
[2], which requires understanding in order to determine whether this degradation is of wider concern, or even the start of a much more serious future concern
[3]. To date, there has been no IGSCC of Alloy 800 in CANDU SG service, although
AECL NUCLEAR REVIEW
5
aecl Nuclear Review
vol 1, Number 1, june 2012
assessing long-term performance of candu out-of-core materials
r.l. tapping, y.c. lu, d.s mancey and z.h. walker
there has been some secondary side pitting detected. Alloy
800 was originally selected for CANDU SGs because of its
predicted excellent resistance to primary side SCC (PWSCC;
primary water SCC), and no susceptibility to PWSCC has
been found in laboratory testing or in service. Figure 1
schematically illustrates the relative behaviour of Alloy 800
and other Ni-base alloys to a range of environments, indicating its suitability for SG applications.
Figure 1
Schematical illustration of the SCC susceptibility of Alloy
800 and other Ni-base alloys in pure water or 0.1% chloride
solution at 350˚C. (Courtesy R.W. Staehle; based on Coriou
data).
6
For some time, the R&D carried out on CANDU SG tube corrosion has addressed all three materials listed in Table 2,
and in some cases Alloy 690TT has also been evaluated for
benchmarking purposes. In the mid-1990’s, it was also decided to use a consistent set of crevice chemistries as the
basis for corrosion testing. These chemistries, outlined
in Table 3, were determined based on hideout return and
MULTEQ calculations from Bruce A SGs, where the crevice
chemistry of the Alloy 600-tubed Bruce A SGs was taken as
the basis because of significant corrosion experienced there
starting in the late 1980’s [4]. Addition of Pb compounds to
these solutions increases the pH slightly. Two significant
aspects dominated the Bruce A SG tubing degradation; one
was the presence of sulphates (with several sulphuric acid
excursions) and consequent acidic crevice chemistries, and
the other was the inadvertent introduction of a lead blanket
into one of the Bruce A SGs, resulting in a significant onset
of IGSCC [5]. Hence there has been focus on the role of both
impurities, and combinations of these with other potentially corrosive impurities, in evaluating the response of all
four materials under plausible CANDU off-specification SG
operating conditions. Much of this work has been carried
out using electrochemical techniques, and results using this
approach will be the emphasis in what follows.
Table 1
Summary of Alloy 800 SG Tubing Service and Associated
Degradation Reports
Unit
(in-service
date)
Country
Degradation
(first detected)
Location
SG Inlet T
(°C)
Cause
Incorrect
Na/PO4 ratio
(wastage);
condenser
leaks (seawater; pitting)
Point Lepreau
(1983)
Canada
Wastage/pitting
(1987)
TTS1 sludge
pile
310
Darlington
(2003,2005 and
2006)
Canada
Shallow pitting
(Reported in 2004)
TSP2 (grids)
309
Biblis A (1974)
Germany
Axial OD
IGA/IGSCC3 (1999;
reported 2005)
Intra tubesheet
(peripheral
tubes)
313
Unterweser
(1979)
Germany
Axial OD
IGA/IGSCC (2005)
TTS and TSP
(grids)
318
Almaraz
(1981/83 inservice; SG
replacements
1996, 1997)
Spain
Circumferential
IGSCC (?); denting
TTS
327
Table 1 summarizes the available in-service Alloy 800 SG
tubing corrosion degradation data. Note that only in-service corrosion-related or corrosion-suspected data are included; there are many more incidents of tube damage and/
or plugging as a consequence of mechanical issues, such as
fretting against supports and loose parts damage. These
data, along with the R&D data to be discussed shortly, indicate that a deterministic assessment of long-term future
performance of Alloy 800 in CANDU service is impractical; rather a probabilistic approach is required that can be
adapted to data as they become available.
2.1 Recent CANDU-Related R&D on Alloy 800
2.2 Lead in combination with impurities
CANDU reactors have a variety of SG tubing materials,
de-pending on their design and age. Table 2 summarizes
these materials, along with a summary of any corrosionrelated degradation experienced in CANDU service. Note
that Indian Pressurized Heavy Water Reactors (PHWRs)
have SGs tubed with both Alloy 400 (earlier PHWRs) and
Alloy 800 (recent 220 MWe PHWRs, 540 MWe PHWRs
and planned 700 MWe PHWRs), and there have been no
reported corrosion-related issues with the Alloy 800 tubing.
For some time, it has been recognized that lead (Pb) and sulphur/sulphates are implicated in the corrosion, and in particular intergranular (IGSCC), of SG tube alloys (for instance,
see reference [6]). Lead is a potential life-limiting impurity
for SGs, and is a secondary side impurity that is present in
most, if not all, SGs. As has been discussed previously [6],
there may be a relatively low threshold concentration for PbSCC of Alloy 600. Both Alloy 800 and Alloy 690 have generally better resistance to such IGSCC than Alloy 600, as typified
Impurities;
crevices
(condenser
leaks)
Impurities;
crevices
(condenser
leaks)
Not confirmed; based
on inspection
data (Xprobe); denting
aecl Nuclear Review
vol 1, Number 1, june 2012
assessing long-term performance of candu out-of-core materials
r.l. tapping, y.c. lu, d.s mancey and z.h. walker
by the excellent in-service performance, and also by laboratory investigations, especially when the evaluations are carried out in realistic off-specification chemistries. Given the
need to assure reactor lives of 60 years or more, and consequently for major components such as SGs, it is necessary to
determine that more resistant materials such as Alloy 690
and Alloy 800 have adequate resistance to PbSCC and other
corrosive secondary side impurities for long service lives.
Table 2
Summary of SG Tube Materials in various CANDU Designs
Unit
Country
Tube Material
Supports
Experience
Limited shallow
pitting (significant
sludge piles)
Pickering A
Canada
Monel 400 (Alloy
400)
Carbon steel lattice
bars
Pickering B
Canada
Monel 400 (Alloy
400)
Carbon steel trefoil
broach plates
Bruce A
Canada
Alloy 600 (slightly
sensitized)
Carbon steel trefoil
broach plates
Bruce B
Canada
Alloy 600 (slightly
sensitized)
Carbon steel trefoil
broach plates
Darlington
Canada
Alloy 800
Stainless steel
lattice bars
Point Lepreau
Canada
Alloy 800
Stainless steel
trefoil broach plates
Gentilly-2
Canada
Alloy 800
Stainless steel
trefoil broach plates
No corrosion
Carbon steel trefoil
broach plates
Pitting of a few
tubes; FAC of
support plates
Embalse
Argentina
Alloy 800
Cernavoda 1,2
Romania
Alloy 800
Wolsong-1
South Korea
Alloy 800
Wolsong 2,3,4
South Korea
Alloy 800
Qinshan 3-1,2
China
Alloy 800
Stainless steel
lattice bars
Alloy 600 formed
lattice bars
Stainless steel
lattice bars
Stainless steel
lattice bars
Significant pitting
under deposits
IGA, IGSCC,
fatigue (early
failures of a few
outer row tubes)
Minor IGA in one
unit; FAC of
support plates
Shallow pitting in a
few tubes near 5th
hot leg supports
Pitting and wastage
(phosphate chemistry until 2000) of
some tubes in
sludge pile areas
and whether sulphate could act to reduce the corrosivity of
Pb, for example by complexing the Pb as an insoluble and
non-corrosive complex. As shown in Table 5, sulphate does
not reduce the corrosivity of Pb to SG tubing. In fact, reduced sulphates, thiosulphates and tetrathionates, for example, may be the most aggressive form of sulphur found
on the secondary side of SGs. The reduction of sulphate to
sulphide during operation is well-known [6], as is the formation of intermediate oxidation state sulphur (IOSS) such
as thiosulphate and tetrathionate following re-exposure to
aerated or oxidizing conditions during SG layup, but only recently has it been shown clearly that these chemistries can
cause severe corrosion in service as a consequence of typical, although not optimal, operating conditions [5]. Given
the widespread presence of sulphates in SGs, it seems likely
that “reduced sulphate” or more precisely IOSS degradation
is more widespread than has been reported.
No corrosion
No corrosion
No corrosion
No corrosion
The use of highly concentrated alkaline or acidic solutions to simulate potentially corrosive crevice chemistries
can produce overly conservative and possibly misleading
results [7]. Preferable is the use of crevice chemistries
that simulate the range that can be achieved within the
experience base of chemistry upsets while operating on
all volatile treatment (AVT) secondary side chemistry, as
summarized in Table 3. Thus Alloy 800 SG tube samples,
along with other SG tubing materials, were exposed to the
chemistries outlined in Table 3 with the addition of various
secondary side impurities, with a focus on Pb but looking
for synergistic corrosive effects. Figure 2 illustrates some of
the electrochemical data, which are assessed on the basis of
passivity breakdown of the alloy, which is a measure of the
susceptibility to corrosion under the test conditions. Parallel autoclave tests were used to assess the SCC susceptibility
under the same conditions. The results have been summarized in Table 4, and are also compiled into “recommended”
pH vs. potential zones as illustrated in Figure 3 [1].
As part of this study the role of sulphates and reduced sulphates (thiosulphates, for example) was examined, along
with examination of the interaction of Pb with sulphates,
Figure 2
Potentiodynamic polarization curves of Alloy 800 SG tubing obtained under simulated neutral SG secondary side
crevice chemistry conditions in the presence and absence
of 500 ppm PbO contamination.
Table 3
Compositions of Simulated Crevice Solutions used for CANDU SG Tube Corrosion Evaluations
Simulated Crevice
Environment
pH at temperature
(°C)
“Neutral” crevice
(“reference solution”)
pH300 6.10
pH150 6.03
Alkaline crevice
pH300 9.26
pH150 10.10
Acidic crevice
pH300 3.22
pH150 2.32
Solution
Composition*
0.15M Na2SO4
0.3M NaCl
0.05M KCl
0.15M CaCl2
Add 0.4M NaOH to
“reference solution”
Add 0.05M NaHSO4 to “reference
solution”
Neutral pH at T for
this composition
pH300 5.16
pH150 5.56
pH300 5.14
pH150 5.56
pH300 5.16
pH150 5.56
Figure 3 shows that Alloy 800 is stable in a range of secondary side chemistries, and should be corrosion resistant
AECL NUCLEAR REVIEW
7
aecl Nuclear Review
vol 1, Number 1, june 2012
assessing long-term performance of candu out-of-core materials
r.l. tapping, y.c. lu, d.s mancey and z.h. walker
to chemistry transients that result in crevice chemistries
within this range. What remains to be determined is the
allowable time to remain outside the range, and the effect
of multiple transients in and out of the range. These are
conditions that can occur during operation, and extended
service lives imply that these conditions might occur more
often. Both Alloy 800 and Alloy 690 have much larger safe
ECP and crevice chemistry ranges than Alloy 600 or Alloy
400, and thus provide more protection against transients.
It remains to be determined if there is a cumulative effect of
transients that can affect SG tubing behaviour beyond that
known from service experience to date, or from laboratory
testing [6]. What also remains to be determined is the effect
of cold work or surface damage on long term behaviour. Recent examinations of Alloy 800 tubing removed from a CANDU SG showed shallow pitting in an area with possible cold
work [9]. Cold work can occur, for example, where tubes
intersect support structures, as well as from fabrication defects, so cold work impacts on tube life could be significant
over long service lives.
stations with service life spanning from two to 27 years was
examined as part of a material aging assessment. It was not
possible to find archived materials that matched the heats
of materials removed from service. Table 5 summarizes the
ex-service tubing assessed and results of the tests. No evidence of an aging effect was found, although it was noted
that the aged materials appeared to have more boron on the
grain boundaries than the archived materials. It is not clear
that this latter result has any significance [10].
In order to better predict future behaviour of SG component life, and in particular tube life, R&D is also underway
on probabilistic methodologies to make these predictions.
Until recently, as noted earlier, Alloy 800 SG tubing had not
been reported to have experienced any significant corrosion in service times up to 30 years or more. Although this
is still the case for CANDU SGs, recent European experience
shows that Alloy 800 can experience significant in-service
degradation under certain conditions. This in-service database, along with laboratory data, can be used to develop
a Bayesian model for Alloy 800 tube performance over long
times, the “degradation free lifetime”. The details will not
be reproduced here [11], but the conclusion is that Alloy
800 should perform reliably for at least 60 years based on
the current data. It should be remembered that most of the
laboratory data, as noted above, were generated using static (in time and concentration) tests under various chemistries, but not under transient or oscillating conditions. Thus,
work is continuing to address the possible consequences of
the accumulation of minor chemistry transients on the long
term stability of Alloy 800.
Table 4
Impurity Mixes Used to Evaluate Susceptibility of SG Tubing to Secondary Side Corrosion
Impurity Mix
Figure 3
Recommended ECP/pH zone with minimized corrosion for
SG alloys at CANDU SG operating temperature.
8
Recent examinations of tubing removed from one Canadian
CANDU SG suggested that an ageing effect might be increasing the corrosion susceptibility of the “aged” (ex-service)
tubing, based on results from electrochemical tests, such
as those used to develop the data shown in Figure 3. The
major difference between the aged material and other materials was that the Ti/C ratio of the ex-service material was
very high, approximately 28 compared to the specification
of >12, although this would normally be predicted to have
a positive effect. As a consequence of this finding, a series
of tests was carried out on archived and ex-service tubing
to determine if there was any evidence for an aging effect.
Steam generator (SG) tubing removed from three CANDU
PbSO4
PbCl and PbS
PbO + SiO2
Al, PbO + Al
Cu, PbO +
Cu
Mg, PbO+Mg
Effect on Alloy 800
No evidence SO4 could immobilize Pb and reduces effect of Pb
on SG tube degradation.
More aggressive than PbO to all SG tube alloys
Silica appears to inhibit Pb-induced degradation of SG tube
alloys
Al impurities apparently have no effect on corrosion
Cu contamination increases ECP and pitting susceptibility
(oxidant effect)
Mg increases ECP and decreases pitting potential. The effect of
Mg on PbSCC requires further investigation.
3. Carbon steel corrosion
CANDU reactors have used carbon steel (A106 Grade B) for
PHTS piping that connects the fuel channels to the rest of
the PHTS circuits. Some time ago it was recognized that
magnetite fouling of the primary side of CANDU SG tubing
was a consequence of the corrosion of, and deposition of
corrosion products from, carbon steels. Subsequently it
was recognized that this corrosion product deposition was
assessing long-term performance of candu out-of-core materials
r.l. tapping, y.c. lu, d.s mancey and z.h. walker
aecl Nuclear Review
vol 1, Number 1, june 2012
largely a consequence of flow accelerated corrosion (FAC)
of corrosion products from, carbon steels. Subsequently, it
was recognized that this corrosion product deposition was
largely a consequence of flow accelerated corrosion (FAC)
of the outlet feeder piping (“feeders”), and was exacerbated
by unusually low Cr contents of the steel and local hydraulic
behaviour. Similarly, several cases of unexpectedly high FAC
rates of carbon steel have occurred in the secondary side of
CANDU SGs, where fouling-induced hydraulic disturbances
have resulted in unanticipated damage. Again, the role of Cr
has been a factor in determining the pattern of damage. In
terms of predicting carbon steel component life, these factors need to be included, although they were not at the time
of initial design. Finally, the intergranular cracking found in
some outlet feeders at one CANDU station (PLGS), although
apparently limited to just 13 feeders at the one station [12],
and a repaired feeder-to-feeder weld, remains a concern
given the several thousand similar feeder pipes in service
in other reactors and, to date, no unambiguous mechanistic
interpretation of the cracking.
Table 5
Ex-Service Alloy 800 Tubing Examined and the Results
Summarized
STATION
DNGS
DNGS
CNG2
PLGS
Ex Service Tube ID
D4 SG3 R54C76
D4 SG1 R49C61
G2 SG3 X69Y54
PL SG3 R32C65
Removed
1995
2003
Apr. 2009
Dec. 2009
In-Service Date
Jun. 1993
Jun. 1993
Oct. 1983
Feb. 1983
Calendar Year
2
10
25.5
26.8
EFPY/Hot Years
1.6/1.7
8.4/8.7
20.4/--†
20.94/21.99
Surface Chromium
No depletion was
detected
No depletion was
detected
No depletion was
detected
No depletion was
detected
Electrochemical
corrosion behaviour
Identical to archived
new tubing
Identical to archived
new tubing
Identical to archived
new tubing
Identical to archived
new tubing
Chemical compositions
Met with ASME
SB-163 Standard
Met with ASME
SB-163 standard
Met with ASME
SB-163 Standard
Met with ASME
SB-163 Standard
Hardness (HV)
160
165
163
173
Grain Size (μm)
13.8/15.0
11.0
8.2/8.8
9.4/8.5
Cr depletion at grain
boundaries
Not detected
Not detected
Not detected
Not detected
Boron segregation at
grain boundaries
-
Observed
Observed
Significant
Feeder piping corrosion
General corrosion of carbon steel under the pH ~10 chemistry of a CANDU PHTS is low enough that long-term acceptable performance is expected. However, although FAC rates
under CANDU PHTS operating conditions are very low on
average, the FAC is localized to specific areas of the outlet
piping where rates of wall loss up to 175 µm/year have been
measured. Given that the feeder pipes have wall thicknesses
of ~8 mm, these wear rates have a significant impact on feeder life, in some cases requiring premature replacement [12].
Figure 3 provides an overview of the feeder pipes that have
been replaced in Canadian CANDU reactors as a consequence
of wall thinning. As has been discussed previously [14],
the FAC mechanism responsible for the wall thinning found
in CANDU outlet feeders is primarily a consequence of the
flow of Fe-unsaturated water into the outlet feeders. The
PHTS coolant becomes significantly unsaturated in Fe as it
traverses the non-ferrous zirconium alloy pressure tubes
and is heated to ~300°C in the process, thus leaving the
water able to dissolve more Fe [15]. This, combined with
turbulent flow and high mass transfer as the coolant enters
the typically tight radius bends of the outlet feeder piping,
results in wall-thinning rates high enough to reduce feeder
pipe life significantly (although note that these rates are
low compared to those found at lower temperatures in
feedwater piping). Based on available solubility data for Fe
and Fe3O4 some reduction in rate can be achieved by lowering the pH to ~10.2 to 10.4 from the typical range of 10.2
to 10.8, and experiments are underway to determine if the
pH can be lowered further and achieve lower FAC rates.
Chemistry modifications are of value to mitigate feeder
thinning in operating reactors. For new reactors, and for
replaced feeders, for instance during refurbishments for
extended life, it has been shown that specifying Cr contents
of the steel to be ≥0.3 wt % reduces feeder thinning rates
by 50%, demonstrated by both in-service performance and
in- and out-reactor laboratory testing [16]. Typically the Cr
contents of early CANDU reactor feeder pipes are very low,
<0.02 wt % for instance, which is a consequence of the low
Co specification originally imposed in order to reduce corrosion product activation. More recent reactors, and all new
CANDU-6 reactors, specify higher Cr steel, while remaining
within the A106 specification. Given that the feeder pipe
FAC mechanism is well understood and predictable [16],
this degradation mechanism can be readily managed for existing and extended design lives.
Laboratory tests using AECL’s FAC loop have demonstrated
that the production of hydrogen by the FAC reaction can be
completely suppressed by the presence of very low concentrations of dissolved oxygen, and that this suppression
occurs without significant changes in the rate of FAC. In
these experimental loop tests, abrupt changes in dissolved
oxygen concentration were made to stop and start the entry of atomic hydrogen into the test samples and hydrogen
effusion probes used to measure the diffusion of hydrogen
through carbon steel [17].
Of more concern than FAC, because there is not as good an
understanding of the mechanistic details, has been feeder
cracking. Although all of the early CANDU-6 reactors have
similarly fabricated feeder pipes, only those at Point Lepreau Generating Station (PLGS) have experienced cracking.
Figure 4 shows the chronology of the detected cracks. The
cracking experienced at PLGS is intergranular stress corrosion cracking, both from the inside and outside surfaces
with no obvious grain boundary segregation of impurities. The feeder bends at these early reactors were heavily
AECL NUCLEAR REVIEW
9
assessing long-term performance of candu out-of-core materials
r.l. tapping, y.c. lu, d.s mancey and z.h. walker
aecl Nuclear Review
vol 1, Number 1, june 2012
cold-worked during fabrication [12]. The most plausible
explanation for this cracking is that low temperature creep
cracking of highly cold-worked material, combined with
FAC-induced hydrogen from the inside surface, are contributing to the cracking [18]. To test this hypothesis, a number of experiments have been carried out in the FAC test rig
with pre-cracked test spool pieces subjected to an applied
load. In these experiments, the FAC process can be turned
14
PLGS (1) S08A;
13
Total Number of Replacements
12
PLGS (3) C13A, N19A, P09A;
G-2 (1) G09A Weld;
11
10
9
PLGS (1) N12C;
PLGS (3) K16A, Q08A, U15C;
8
PLGS (2) H12A, N16C;
7
6
5
PLGS (1) D14A + 6 False
Calls (P18C, O07C, L16C,
E08C, E14C, K05C);
4
3
2
PLGS (1) N11A;
1
0
2010
2009
2008
2007
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
Figure 4
Historical Record of Point Lepreau Outlet Feeder Cracking.
on and off by modifying the oxygen content, and thus the
role of FAC-induced hydrogen can be determined. Results
to date suggest that FAC hydrogen is a contributor to the IGSCC; with FAC-induced hydrogen present, the crack growth
rate is significantly faster than when hydrogen generation is
suppressed (but FAC continues) [17].
Separate tests have demonstrated that low temperature
creep of A106B carbon steel is measurable, with crack
growth rates comparable at 300°C to those measured for
feeder cracks at PLGS, for heavily cold-worked material
stressed close to yield [18].
The role of operating transients on feeder cracking has not
been investigated. What remains to be understood is why
the PLGS IGSCC has not occurred in feeder piping elsewhere
(although IGSCC was observed in a repaired feeder weld at
Gentilly-2 NGS). For more recent reactors, the feeder bends
are stress relieved and thus are not expected to be at significant risk of cracking.
4. Other systems
10
Most of the secondary side of power plants is constructed
with carbon steel piping. Recently, after the discovery of
the ubiquitous nature of FAC under feedwater conditions,
FAC-resistant materials (Cr-Mo alloys; stainless steels) are
used in high susceptibility areas such as susceptible geometries, low pH, two-phase flow, and flows with Fe concentrations less than saturation, such as moisture separator reheater piping [19] are used. FAC of secondary side
systems and piping can be engineered out with appropriate materials selection, designed to avoid turbulence and
impingement, and appropriate operating chemistry. For
older plants, inspection is a key part of the management approach, and at CANDU plants, the CHECWORKS code of the
Electric Power research Institute (EPRI) is applied. AECL
has also applied this code, along with CFD (computational
fluid dynamics) calculations, to assess FAC of carbon steel
components on the secondary side of SGs. There have been
several instances of secondary side FAC of key secondary
side SG components, as noted in Table 2. R&D continues
on improving the predictability of CHECWORKS in terms
of quantitation of expected wall thinning rates at high risk
locations, using accumulated plant data as a baseline for
predictions, along with improved understanding of the key
factors influencing FAC of carbon steel. For buried piping, the use of non-metallic materials is being considered,
and modelling of service water piping systems has shown
where conditions particularly susceptible to microbiological corrosion, or microbiological corrosion (MIC) (for instance, vertical deadlegs), can be designed out. In addition,
the R&D has provided the data needed for improving operating specifications to minimize the ingress of nutrients
that can lead to biological growth and subsequent MIC.
Low temperature piping, typical of that used in service water systems, is also susceptible to localized corrosion, either
under deposits (tuberculation) or as a consequence of MIC
attack. In collaboration with EPRI, AECL has been developing models of tuberculation and MIC to incorporate into
EPRI’s BPWORKS code, which is used to identify high risk
areas where inspections of piping should be carried out.
This is particularly important for buried piping, where the
consequences of undetected degradation can be significant,
and the cost of carrying out unnecessary inspections is also
a major concern. Thus, R&D in this area is focused on inspection technology developments, along with improved
risk ranking techniques.
Other systems of concern for long life, and where CANDUrelated R&D is being carried out, include concrete structures, cables and the calandria vessel. The main issue with
concrete is that many concrete structures must remain intact and act as a barrier to prevent release of nuclear materials for very long times. Concrete degradation mechanisms for unirradiated structures are well known, although
work remains to be done on irradiated concrete. The key
life management issue for concrete is improving inspection
technologies such that non-destructive methods of assessing concrete structure condition can be quantified. Models for the degradation of concrete over long times may
need to be developed to relate inspection findings to life
aecl Nuclear Review
vol 1, Number 1, june 2012
assessing long-term performance of candu out-of-core materials
r.l. tapping, y.c. lu, d.s mancey and z.h. walker
prediction. For cables, aging models have been developed,
and related to inspection results from various destructive
and non-destructive techniques. To date, experience with
cables in various CANDU reactors, and in AECL’s NRU reactor, all show that the cables are in good condition and
can remain in service. Finally, the calandria vessel, which
contains the heavy-water moderator, at 60 to 70°C, which
surrounds the pressure tubes, requires some assessment of
remaining life. This vessel experiences a neutron flux up
to 2x1021 n/cm2, and is fabricated from type 304L stainless steel. There are essentially no data available showing
the long-term effects of neutron irradiation of this material
50
Wall Thinning Greater than Expected
DNGS-1 (3) H12W, Q24W, P10E;
PNGS-8 (2) B14W, N21W;
45
BNGS-3 (1) A15E;
DNGS-2 (2) M01E, L01W;
G-2 (1) C05C;
PNGS-1 (3) C08E, C15W, E19W;
Total Number of Replacements
40
35
G-2 (1) D06C;
DNGS-1 (3) C07W, J01W, O24W;
30
25
PLGS (6) C06C, C17A, D05C,
D18A, E19A, H02A;
G-2 (1) G09A;
PNGS-4 (2) B11E, B12W;
20
BNGS-3 (2) A14W, A16W
BNGS-6 (4) A09E, A11E,
A14W, A16W
15
PNGS-1 (4) B09E, B11E, B12W, C07W;
10
BNGS-6 (3) A10W, A17E, A15E;
BNGS-8 (2) F13W, L05E;
DNGS-3 (1) O01E;
PNGS-4 (3) C06E, C16E, E07W;
5
0
2011
2010
2009
2008
2007
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
Figure 5
Historical record of outlet feeder replacements in CANDU
reactors because of feeder thinning.
at low temperatures, but several analyses using extrapolations from available data indicate that the vessel and vessel
welds would not have experienced any significant irradiation-induced effects after 30 years of service, and can remain in service. To quantify this, it is proposed to destructively examine other structures from inside the calandria
vessel in order to assess any irradiation effects.
5. Conclusion
Materials performance remains the primary long-term
risk for most industrial systems, and especially for nuclear
power plants. Predicting that performance quantitatively
has always posed challenges [20], and an even bigger challenge is to predict performance for extended reactor lives of
60 to 80 years. For CANDU reactors, life extension implies
replacement of the fuel channel assemblies, and hence the
materials issues relate to whether or not out-reactor components and systems need to be replaced. In most cases, inservice experience and laboratory data indicated that passive components such as steam generators, piping, etc. do
not need replacement. Significant exceptions are where Alloy 600 and Alloy 400 were used for used for SG tubing, and
where carbon steel was used for secondary side SG support
structures. The most recent CANDU reactors, built since the
1980’s, and those being refurbished now, use designs, components and materials, in addition to appropriate operating and maintenance strategies, based on results from R&D
programs carried out at AECL’s laboratories. These results
indicate that Alloy 800 SG tubing, Cr-enriched carbon steels
and appropriately designed service water systems can provide low-risk long term service.
References
[1] H. Coriou, L. Grall, P. Olivier and H. Willermoz, 1968, “Influence of Carbon and Nickel Content on Stress Corrosion Cracking of Austenitic Stainless Alloys in Pure or Chlorinated Water at 350 C”,
in Fundamental Aspects of Stress Corrosion Cracking; NACE-1, NACE, Houston, pp. 352-359
[2] J. Slade, V. Moroney, J. Gorman and T.S. Gendron, 2007, “A CANDU Utility Perspective on Using World Experience to Manage Alloy 800 SG Tube Degradation”, Proceedings 13th International
Conference on Environmental Degradation of Materials in Nuclear Power Systems, Whistler, B.C., Canada
[3] R. Kilian, J. Beck, H. Lang, T. Schönherr, and Martin Widera, September 2010, “Root Cause Analysis of SG Tube ODSCC Indications within the Tube Sheets of NPP Biblis Unit A”, Proceedings of the
International Symposium, Fontevraud-7, Avignon, France, Paper A154-T06
[4] Y.C. Lu, 2010, “Effect of Hazardous Impurities on Steam Generator Tube Degradation”, Proceedings ICONE-18, Xi’an, China, Paper ICONE18-30120
[5] K. Sedman and D. Durance, November 2009, “Update of the SG Tube Intergranular Attack/Stress Corrosion Cracking in Bruce Unit 4, Proceedings 6th CNS International Steam Generator Conference, Toronto
[6] R.W. Staehle and J.A. Gorman, 2004, “Quantitative Assessment of Sub-modes of Stress Corrosion Cracking on the Secondary Side of Steam Generator Tubing in Pressurized Water Reactors”, Corrosion, Part 1: 59 (2003) p.931-994; Part 2: 60 (2004) p.5-63; Part 3: 60, pp.115-180
[7] P.V. Balakrishnan, S.M. Pagan, A.M. McKay and F. Gonzalez, 1995, “Hideout and Hideout Return: Laboratory Studies and Plant Measurements”, Control of Corrosion on the Secondary Side of
Steam Generators, Airlie, Virginia, NACE, pp. 683-708
[8] Y.C. Lu, 2007, “Define Optimal Conditions for Steam Generator Tube Integrity and an Extended Steam Generator Service Life”, Proceedings ICONE-15, Nagoya, Japan, Paper ICONE15-10854
[9] Y.C. Lu, G. Goszczynski and S. Ramamurthy, July 2009, “Degradation of Alloy 800 under Steam Generator Secondary Side Crevice Conditions”, Pro-ceedings of the 17th International Conference
on Nuclear Engineering, ICONE-17, Brussels, Belgium, Paper ICONE17-75695
[10] Y. Lu, S. Ramamurthy, G. Goszczynski, 2012, “An Aging Assessment on Ex-Service Alloy 800 Steam Generator Tubing”, Nuclear Engineering and Design, 242, pp. 91-99
[11] Y.C. Lu, R.L. Tapping and M.D. Pandey, November 2009, “Degradation of Alloy 800 Steam Generator Tubing and its Long-Term Behaviour Predictions for Plant Life management”, Proceedings
6th CNS International Steam Generator Conference, Toronto
[12] J.P. Slade and T.S. Gendron, August 2005, “Flow Accelerated Corrosion and Cracking of Carbon Steel Piping in Primary Water – Operating Experience at the Point Lepreau Generating Station”,
Proceedings 12th International Conference on Environmental Degradation of Materials in Nuclear Power Systems – Water Reactors, Salt Lake City, U.S.A., pp. 773-784
[13] Z.H. Walker, July 2004, “Managing Flow Accelerated Corrosion in Carbon Steel Piping in Nuclear Plants”, Proceedings of ASME PVP Conference (PVP2004), San Diego, U.S.A., Paper PVP20042251
[14] R.L. Tapping, 2008, “Materials Performance in CANDU Reactors: The First 30 years and the Prognosis for Life Extension and New Designs”, Journal of Nuclear Materials, 383, pp. 1-8
[15] D.A. Guzonas and L. Qiu, October 2004, “A Predictive Model for Radio-nuclide Deposition Around the CANDU Heat Transport System”, International Conference on Water Chemistry of Nuclear
Reactor Systems, San Francisco, U.S.A., pp.2079-2086
[16] Z.H. Walker, A.J. Elliot, D.S. Mancey and B. Rankin, October 2006 , “Resistance of SA-106 Carbon Steel containing >0.30 wt% Cr to Flow Accelerated Corrosion Under CANDU Reactor Outlet
Feeder Pipe Conditions”, International conference on Water Chemistry of Nuclear Reactor Systems, Jeju Island, Korea
[17] D.S. Mancey, AECL, unpublished results
[18] M.D. Wright, AECL, unpublished results
[19] B. Chexal, et al., 1998, “Flow-Accelerated Corrosion in Power Plants”, EPRI Report TR 106611-R1
[20] P.L. Andresen, T. Angeliu and L.M. Young, February 11-15, 2001, “Immunity, Thresholds and Other SCC Fiction”, in Chemistry and Electrochemistry of Corrosion and Stress Corrosion Cracking:
A Symposium Honouring the Contributions of R.W. Staehle, edited by R.H. Jones, TMS, New Orleans, Louisiana, pp. 65-82.
AECL NUCLEAR REVIEW
11
12
FULL ARTICLE
Abstract
The localized corrosion resistance of
nuclear-grade Alloy 800, which is one of
the preferred steam generator (SG) heat
exchange tube materials of CANDU and
PWR reactors, was studied under simulated
SG secondary side crevice chemistry
conditions at ambient temperature as well
as at elevated temperatures. Series of cyclic
potentiodynamic polarization tests were
performed to study the localized corrosion
resistance of Alloy 800 as a function of
chloride ion concentration in the SG crevice
solution at 40˚C, 150˚C and 300˚C. Based
on the experimental results, empirical
equations were provided for calculating
the pitting potential of nuclear grade
Alloy 800 in the SG secondary side crevice
chemistries with different levels of chloride
concentration at SG layup, startup and
operating temperatures.
localized corrosion of nuclear
grade alloy 800 under steam
generator layup, startup and
operating conditions
Y. Lu*
Atomic Energy of Canada Limited, Chalk River Laboratories, Chalk River, Ontario, Canada, K0J 1J0
Article Info
Keywords: steam generator, alloy 800, localized corrosion, potentiodynamic polarization, crevice corrosion
Article history: Received 19 March 2012, Accepted 25 June 2012, Available online 30 June 2012.
*Corresponding Author: (613) 584-3311 ext. 43258, [email protected]
Introduction
Significant corrosion degradation of steam generator (SG) tubing has been found
in the SG secondary side worldwide. Pitting corrosion and crevice/underdeposit corrosion, intergranular attack (IGA) and stress corrosion cracking (SCC)
are among the possible major forms of degradation for SG tube materials. The
impurities can hide out and concentrate in SG crevices, leading to an aggressive
local chemistry. Corrosion-related SG tube degradation may take place in tube/
tube-support crevices, tube/tubesheet crevices or under sludge deposits when the
secondary side environment contains oxidants and/or is contaminated with chloride. Experimental data suggest that all SG tube materials can be susceptible to
corrosion degradation under some particular off specification SG crevice chemistry conditions [1]. The tolerance to the chemistry upset for each SG tube alloy is
different. In recent years, AECL has been working proactively to deal with the SG
degradation issues. An effective way to minimize the SG material degradation is
through effective water chemistry control. To support the proactive materials degradation management (PMDM) of SG materials, high temperature electrochemical
tests were performed under plausible SG crevice chemistry conditions for all SG
tube alloys. This knowledge has been documented in the form of a recommended
ECP/pH zone that can be incorporated into a program for SG system monitoring
[1, 2]. Tests, especially simulated accelerated corrosion tests and SCC tests, have
been performed to verify the boundary leading to SG tube degradation under transient and off-specification conditions [3]. The recommended ECP/pH zone could
be used as a tool for safeguarding the integrity of SG tubing, with the expectation that operation in these “safe” zones will assure at least 60 years of service.
It should be noted that the plausible crevice chemistry conditions, which were
used to develop the recommended ECP/pH maps of different SG alloys, were
based on formulations suggested by Pagan used in previous work [4] and later
verified by Burton and Turner [5]. However, based on an average of all Darlington Nuclear Generating Station hideout return surveys since 2001, McKay and
Pagan found that the SG crevice solution may not always be as aggressive as previously suggested [6]. It is therefore necessary to evaluate how the boundaries
of the recommended zones change as a function of the concentrations of SG impurities. Chloride is one of the major SG impurities that cause the break down
of the passive film and initiate localized corrosion in passive metals, including
SG tubing materials [1]. Work has been performed by AECL to study the impact of chloride ion concentration on the upper boundary of the safe ECP/pH
zone leading to SG tube alloy degradation. The overall objective of this task is
to evaluate the effect of chloride concentration on the SG tube corrosion degradation. The results will provide information to determine how the safe ECP/pH
AECL NUCLEAR REVIEW
13
aecl Nuclear Review
vol 1, Number 1, june 2012
zone boundaries move with variation of SG impurity concentrations at the same high temperature pH. Acidic crevice chemistry is excluded from the recommended ECP/
pH zone and according to a study by Balakrishnan CANDU
SG crevice chemistry is normally in the neutral pH range
even under condenser leakage conditions [7]. Therefore,
the effect of chloride ion concentration on the corrosion
of SG tubing was focused on neutral crevice chemistry
conditions. This paper summarizes the high temperature
electrochemical experimental results from these studies.
1. Experimental Conditions
1.1 Test Material
Samples for electrochemical measurements were 10 mm
long segments cut from ASTM B 163 standard seamless
heat exchanger tubes. In this paper, only data from Alloy
800 SG tubing will be presented. Nuclear grade Alloy 800
SG tubing tested was from Noranda Metal Industries Ltd.,
where the final heat treatment was a bright annealing at
980 ± 10°C. The outer diameter (OD) surface of the test
specimens was finished by grinding with 600 grit silicon
carbide paper and ultrasonically cleaning first with acetone, and then with ethanol before the tests. The detailed
information on the tube material is listed in Table 1.
Table 1
Detailed Information on the SG Tube Materials for Tests
(Based on Mill Test Certificates)
Alloy
Heat No.
800
HH9043A
Composition wt %
Size
C
Si
Mn
P
S
Cr
Ni
Co
Ti
Cu
Al
N
Fe
0.625 x 0.044” Wall
0.015 0.10 0.80 0.009 0.002 21.70 34.11 0.012 0.42 0.03 0.41 0.028 42.41
(15.88 x 1.12 mm)
1.2 Test Environments
The corrosion susceptibility of Alloy 800 SG tubing material
was evaluated by performing a series of electrochemical
measurements under the SG crevice conditions shown in
Table 2 through Table 4. The crevice solutions were determined by referring to simulated representative CANDU SG
crevice environments based on the systematic SG hideout
return results from CANDU Nuclear Generating stations,
and on analysis of the intact tube/bar crevice deposits
[4]. To study the effect of the chloride concentration on
the corrosion degradation of Alloy 800 SG tubing, different amounts of chloride were added to the neutral crevice
chemistry by meeting the following criteria:
14
1. The total concentration of sulphate in the crevice solution
should remain unchanged (or the changes are negligible).
2. The pH of the solution at temperature should be maintained at, or at least close to, a fixed value by using pH
localized corrosion of nuclear grade alloy 800 under steam
generator layup, startup and operating conditions - y. lu
adjusters, such as 10-4 M NaOH or HCl.
The calculation of the at-temperature pH values for the selected crevice solutions were performed using the OLI software, Version 2.048, from OLI Systems Inc. (NJ, USA). The
detailed method and theories used for the calculation in
Stream Analyzer can be found in reference [8] and its citations. Stream Analyzer has two embedded activity frameworks. The one selected for this report was the Aqueous H+
Framework [8].
In this work, only the bulk chloride concentration of the solution is considered. Chloride concentration gradient on the
alloy surface is not significant if Alloy 800 is passive. Under
such conditions, the passive current density is so low that
the potential gradient in the solution near the electrode
surface can hardly result in any significant chloride ion gradients near the passive surface.
Table 2
Neutral Crevice Solution with Different Chloride Concentrations for Tests at 300°C (pH 300°C = 6.10) Unit: mol/kg
[Cl-]
Na2SO4
NaCl
KCl
CaCl2
HCl
0.01
0.1500
0.0027
0.0005
0.0014
4.15E-03
3.98E-03
NaOH
0.1
0.1500
0.0443
0.0074
0.0222
0.65
0.1500
0.3000
0.0500
0.1500
1
0.1500
0.4615
0.0769
0.2308
2.2E-04
4
0.1500
1.8462
0.3077
0.9231
6.2E-04
Table 3
Solutions for Tests Performed at 150°C (pH 150°C = 6.03)
Unit: mol/kg
[Cl]
Na2SO4
NaCl
KCl
CaCl2
HCl
0.01
0.15
4.57940x10-3
7.63000 x10-4
2.29000 x10-3
8.29711 x10-5
8.41592 x10-5
NaOH
0.1
0.15
0.0461173
7.68600 x10-3
0.0230590
1
0.15
0.461538
0.0769230
0.230769
1.10203 x10-5
4
0.15
1.846154
0.307692
0.923077
5.82434 x10-5
Table 4
Solutions for Tests performed at 40°C (pH 40°C = 6.83)
Unit: mol/kg
[Cl]
Na2SO4
NaCl
KCl
CaCl2
HCl
0.01
0.15
4.61568E-03
7.69044E-04
2.30814E-03
5.75180e-7
NaOH
0.1
0.15
0.046154
7.69205E-03
0.023077
4.75957e-7
1
0.15
0.461538
0.0769230
0.230769
1.72694e-7
4
0.15
1.846154
0.307692
0.923077
1.44906e-6
aecl Nuclear Review
vol 1, Number 1, june 2012
2. Electrochemical Tests
Electrochemical methods are usually used to determine the
corrosion susceptibility of SG tube material under different
environments. Potentiodynamic polarization is a technique
where the potential of an electrode is swept at a selected
rate by application of a current through the electrolyte using a potentiostat. Potentiodynamic polarization tests can
provide information on corrosion kinetics (rate of corrosion) as a function of ECP in specific environments. In order
to determine the effect of chloride ion concentration on the
degradation of SG tubing in SG crevices and under deposits, a series of electrochemical tests of Alloy 800 SG tubing
were performed in simulated SG crevice chemistries listed
in Table 2 through Table 4.
All electrochemical polarization data were performed in
static autoclaves using a typical three-electrode system.
The schematic of the electrochemical cell and sample
mounting for high temperature electrochemical measurements is shown in Figure 1. An EG&G Model 263A/99 Potentiostat/Galvanostat with a floating/auxiliary input option is used for the potentiodynamic polarization tests in
autoclave systems. The scan rate is fixed at the ASTM standard recommended rate, 0.167 mV/s [9]. All samples are
tested under isothermal conditions. Internal Ag/AgCl/0.65
M KCl high temperature reference electrodes were used to
make high temperature electrochemical measurements.
Electrochemical measurements at 40°C are performed in a
three-electrode system shown in Figure 2. An oil bath is
used to maintain the test temperature. A saturated calomel electrode (SCE) is used as the reference electrode. A
platinum foil is used as a counter electrode. To minimize
the solution IR drop, the Luggin capillary of the reference
electrode is placed close to the sample surface
( 1 mm).

For this work, all potentials were converted to the standard
hydrogen electrode scale (SHE) [10], [11].
localized corrosion of nuclear grade alloy 800 under steam
generator layup, startup and operating conditions - y. lu
Purging line
Autoclave
Zircaloy Lining
Ag/AgCl/(0.65M KCl)
reference electrode
Sample holder
Tube sample
Pt counter electrode
Figure 1
A three-electrode system for electrochemical measurements in a static autoclave.
The samples for electrochemical tests were 10 mm long
segments cut from SG tubing. The external surface was finished with 600 grit silicon carbide paper and ultrasonically
cleaned first with acetone and then with ethanol before the
tests.
3. Experimental Results
The experiments were performed at three temperatures,
representing SG layup, startup, and full power operation.
Electrochemical polarization tests suggest that Alloy 800
is susceptible to pitting corrosion in the simulated neutral
crevice chemistry and the pitting potential is a function of
chloride concentration.
Figure 1
A glass cell assembly used for cyclic potentiodynamic polarization measurements at 40°C.
AECL NUCLEAR REVIEW
15
localized corrosion of nuclear grade alloy 800 under steam
generator layup, startup and operating conditions - y. lu
3.1 Effect of Different Chloride Concentrations on Alloy
800 SG Tube Degradation at 300°C
The cyclic polarization curves of Alloy 800 SG tubing obtained under neutral crevice chemistry conditions at 300°C
with various chloride concentrations were superimposed
in Figure 3. It is clearly shown that the pitting potential of
Alloy 800 is significantly affected by the chloride concentration in the crevice chemistry. Alloy 800 is free of pitting in
neutral crevice chemistry with a chloride concentration ≤
0.01 mol/kg. The existence or absence of a pitting potential can be identified from the polarization curves. For the
polarization curves in solutions containing chloride at concentrations ≤ 0.01 mol/kg, the current density increased
linearly after the potential was raised above 390 mV, and
it decreased following the same slope during the back scan.
The current density increase for potentials above a potential of 390mV was the result of oxygen evolution. The
specimens did not pit and the surface area of the specimens
remained the same. Therefore, the forward scan and the
back scan almost overlapped. For the rest of the polarization curves shown in Figure 3, for solutions with higher
chloride ion concentrations, there was a sudden increase
in the anodic current density resulting from pitting corrosion. The pitting degradation was not reversible and the
anodic current density remained high during the back scan
until the pits stopped growing at a specific lower potential.
The cathodic branches of the polarization curves should be
the same if the high temperature pH is the same. However,
there were some obvious discrepancies in very dilute chloride solutions and very concentrated chloride solutions indicating the pH calculation provided by the stream analyzer
requires further correction under extreme diluted and concentrated solutions.
SG crevice conditions at 300°C can be approximately fitted
by the following empirical equations:
- - Epit= =(451
(451-1966
-1966[Cl
[Cl
mV 0.01
0.01mol/kg
mol/kg> >[Cl
[Cl
0.20mol/kg
mol/kg (1) Equation
Equatio
Epit
])])mV
] ≤] ≤0.20
- - 2- 2
- Epit= =(-288
(-288+ +56.3/[Cl
56.3/[Cl
0.49/[Cl
] )mV[Cl
[Cl
0.20mol/kg
mol/kg
Epit
] –] –0.49/[Cl
] )mV
] >] >0.20
Equatio
(2) Equation
3.2 Effect of Different Chloride Concentrations on Alloy
800 SG Tube Degradation at 150°C
The cyclic polarization curve of Alloy 800 SG tubing obtained
0.800
E (V vs. SHE)
aecl Nuclear Review
vol 1, Number 1, june 2012
0.600
0.01 m Cl-
0.400
0.10 m Cl-
0.200
0.65 m Cl-
0.000
1.0 m Cl-
-0.200
4.0 m Cl-
-0.400
-0.600
-0.800
-1.000
10-6
10-5
10-4
10-3
10-2
10-1
Current density (A/cm2)
Figure 3
Cyclic polarization curves of Alloy 800 obtained at 300°C
under neutral SG crevice chemistry conditions with different chloride concentrations.
Replicated tests were performed under each test condition.
The pitting potential values obtained at 300°C from different tests performed in solutions with various chloride concentrations are listed in Table 5.
Table 5
Pitting potential of Alloy 800 SG tubing in neutral SG crevice
at 300°C.
Cl- (m)
0.01
0.1
0.65
1.0
4.0
Data 1
450 mV*
168 mV
-162 mV
-233 mV
-270 mV
Data 2
390 mV*
233 mV
-155 mV
-244 mV
-283 mV
Data 3
455 mV*
264 mV
-178 mV
-258 mV
-357 mV
Data 4
-181 mV
-260 mV
Data 5
-167 mV
-
Average
231.7 mV*
221.7 mV
-168.6 mV
-245.0 mV
-292.5 mV
STDEV
36.2 mV*
49.0 mV
10.9 Mv
12.5 Mv
44.0 mV
* Free of pitting corrosion
The pitting potential of Alloy 800 under secondary side
crevice chemistry conditions at 300°C as a function of chloride solution is shown in Figure 4.
16
The pitting potential of Alloy 800 SG tubing under secondary
Figure 4
Pitting potential of Alloy 800 SG tubing material at 300°C
under neutral SG crevice chemistry conditions as a function
of chloride concentration.
aecl Nuclear Review
vol 1, Number 1, june 2012
localized corrosion of nuclear grade alloy 800 under steam
generator layup, startup and operating conditions - y. lu
under neutral crevice chemistry conditions at 150°C for
various chloride concentrations were superimposed and
shown in Figure 5. As observed at 300°C, the pitting potential of Alloy 800 is also significantly affected by the chloride
concentration in the crevice chemistry at 150°C. It is seen
that Alloy 800 is susceptible to pitting corrosion at 150°C
even if the chloride ion concentration is as low as 0.01 mol/
kg.
Replicated tests were performed under each test condition.
The pitting potential values obtained at 150°C from different tests performed in solutions with various chloride concentrations are listed in Table 6.
The pitting potential of Alloy 800 under secondary side
crevice chemistry conditions at 150°C as a function of chloride solution is shown in Figure 6. The pitting potential of
Alloy 800 SG tubing under secondary side SG crevice conditions at 150°C can be approximately fitted by the following
empirical equations:
(3)Equation 1
Epit = (1244 -7544 [Cl-]) mV
0.01 mol/kg ≥ [Cl
] ≤ 0.10 mol/kg
Epit = (1244 -7544 [Cl-]) mV 0.01 mol/kg
≥ [Cl--] ≤ 0.10 mol/kg
- 2
Epit = (-193.6 + 166.1/[Cl
- ] – 9.68/[Cl
- 2 ] )mV [Cl
- ] > 0.10 mol/kg
Epit = (-193.6 + 166.1/[Cl ] – 9.68/[Cl ] )mV [Cl ] > 0.10 mol/kg
1.200
Equation 1
Equation 2
Equation 2
(4)
0.01 m Cl-
1.000
0.800
E (V vs. SHE)
0.600
0.10 m Cl-
0.400
0.200
0.000
1.0 m Cl-
-0.200
4.0 m Cl-
-0.400
-0.600
10-6
10-5
10-4
10-3
10-2
10-1
100
Current density (A/cm2)
Figure 5
Cyclic polarization curves of Alloy 800 obtained at 150°C
under neutral SG crevice chemistry conditions with different chloride concentrations.
Table 6
Pitting potential of Alloy 800 SG tubing in neutral SG crevice
at 150°C.
-
Cl (m)
0.01
0.1
0.65
1.0
4.0
Data 1
Data 2
Data 3
Data 4
Data 5
Average
STDEV
1171 mV
472 mV
54 mV
-72 mV
-162 mV
1172 mV
534 mV
72 mV
-43 mV
-132 mV
1163 mV
493 mV
8 mV
-30 mV
-151 mV
38 mV
-
58 mV
-
1168.7 mV
499.7 mV
46.0 mV
-48.3 mV
-148.3 mV
4.9 mV
31.5 mV
24.5 Mv
21.5 Mv
45.2 mV
Figure 6
Pitting potential of Alloy 800 SG tubing material at 150°C
under neutral SG crevice chemistry conditions as a function
of chloride concentration.
3.3. Effect of Different Chloride Concentrations on Alloy
800 SG Tube Degradation at 40°C
The cyclic polarization curves of Alloy 800 SG tubing obtained under neutral crevice chemistry conditions at 40°C
for various chloride concentrations were superimposed in
Figure 7. The pitting potential of Alloy 800 is decreasing as
the chloride concentration in the crevice chemistry increases. Alloy 800 is immune to pitting potential when chloride
concentrations are equal to or lower than 0.01 mol/kg. A
cathodic loop or shoulders could be seen in the polarization
curves shown in Figure 7 indicating the existence of some
residual dissolved oxygen. This would not affect the pitting
potential measurements.
The pitting potential values obtained at 40°C from replicated tests performed in solutions with various chloride concentrations are listed in Table 7.
The pitting potential of Alloy 800 under secondary side
Table 7
Pitting Potential of Alloy 800 SG Tubing in Neutral SG Crevice at 40°C
Cl- (m)
0.01
0.1
0.65
1.0
4.0
Data 1
*
1129 mV
411 mV
371 mV
199 mV
Data 2
Data 3
Data 4
Data 5
Average
STDEV
*
1058 mV
346 mV
316 mV
203 mV
*
1131 mV
348 mV
309 mV
238 mV
425 mV
-
374 mV
-
*
1106.0 mV
380.8 mV
332.0 mV
213.3 mV
*
41.6 mV
36.1 mV
34.0 mV
21.5 mV
* Free of pitting corrosion
AECL NUCLEAR REVIEW
17
localized corrosion of nuclear grade alloy 800 under steam
generator layup, startup and operating conditions - y. lu
aecl Nuclear Review
vol 1, Number 1, june 2012
crevice chemistry conditions at 40°C as a function of chloride solution is shown in Figure 8 and can be approximately
fitted by the following empirical equation:
is shown in Figure 9. It is seen that the pitting potential of
Alloy 800 at 150°C under neutral crevice chemistry conditions follows a linear relationship with the log value of chloride ion concentration when the chloride ion concentration
(5) Equation 1is less than 1 mol/kg. At 40°C and 300°C the linear relation
Epit = (183 + 141/[Cl-] – 4.87/[Cl-]2)mV [Cl-] > 0.10 mol/kg
between Epit and log CCl- can also be seen in the range of 0.1
1.600
m ≤ CCl- ≥1 m. It is possible that the Epit and Log [Cl-] may
0.01 m Cl
1.400
follow a linear relation. Attempts were made to check the
1.200
linear relation between Epit and log [Cl-] but failed because
0.10 m Cl
1.000
of an unreasonable high calculated activity coefficient value
0.800
at high chloride ion concentrations based on the data ob0.65 m Cl0.600
tained from the commercial OLI software Stream Analyzer
1.0 m Cl
0.400
[8] This observation also suggests that the calculated pH
0.200
value of solutions containing very high chloride concentra0.000
tions using the OLI software Stream Analyzer may have cer4.0 m Cl-0.200
tain deviations.
E (V vs. SHE)
-
-0.400
-0.600
1400
-0.800
-1.000
10-9
10-8
10-7
10-6
10-5
10-4
10-3
10-2
10-1
1200
100
2
Current density (A/cm )
800
E pit (mV vs. SHE)
Figure 7
Cyclic polarization curves of Alloy 800 obtained at 40°C under neutral SG crevice chemistry conditions with different
chloride concentrations.
40°C
150°C
300°C
1000
600
400
200
0
-200
-400
-600
0.01
0.1
1
10
Chloride Concentration (mol/L)
Figure 9
Pitting potential of Alloy 800 SG tubing material at 40°C,
150°C and 300°C under neutral SG crevice chemistry conditions as a function of chloride concentration.
3.4 Pitting Potential of Alloy 800 under Neutral Crevice
Chemistry Conditions as a Function of Temperature
Figure 8
Pitting potential of Alloy 800 SG tubing material at 40°C under neutral SG crevice chemistry conditions as a function of
chloride concentration.
The pitting potential of Alloy 800 SG tubing at three different temperatures under neutral SG crevice chemistry conditions as a function of chloride concentration in log scale
18
The pitting potential of Alloy 800 decreases with temperature under the same crevice chemistry conditions. The
pitting potential of Alloy 800 as a function of temperature
could be fitted with equations shown in Figure 10. In the
figure, the pitting potential of Alloy 800 under neutral SG
crevice chemistries with different chloride ion concentrations as a function of temperature in Kelvin scale was plotted against K-1. It appears that the pitting potential in a
crevice solution with fixed chloride concentration follows
a linear relation with 1/K. This observation suggests that
the breakdown of the passivity and the initiation of pitting
corrosion by chloride ions are thermally activated.
aecl Nuclear Review
vol 1, Number 1, june 2012
Temperature
1400
1200
localized corrosion of nuclear grade alloy 800 under steam
generator layup, startup and operating conditions - y. lu
300°C
40°C
150°C
0.1M Cl-
Epit = (451 -1966 [Cl-]) mV 0.01 mol/kg > [Cl-] ≤ 0.20 mol/kg
- 2
EpitE=pit(451
-1966
[Cl-]) mV
0.01 mol/kg
> [Cl-] ≤> 0.20 mol/kg
= (-288
+ 56.3/[Cl
] – 0.49/[Cl
] )mV
- 2
Epit = (-288 + 56.3/[Cl ] – 0.49/[Cl ] )mV [Cl-] > 0.20 mol/kg
(6)
(7)
1000
E pit (mV vs. SHE)
800
5
600
-1
Epit = -892 +6.17x10 K
0.65M Cl-
400
200
1.0M Cl5
4.0M Cl-
-1
Epit = -840 + 3.80x10 K
0
5
-1
Epit = -964 + 4.01x10 K
-200
5
-1
Epit = -937 + 3.54x10 K
-400
-600
0.0016
0.0018
0.0020
0.0022
0.0024
0.0026
0.0028
0.0030
0.0032
0.0034
1/K
Figure 10
Pitting potential of Alloy 800 SG tubing material under neutral SG crevice chemistry conditions with different chloride
ion concentrations as a function of 1/K.
4. Conclusions
Cyclic potentiodynamic polarization measurements were
performed on Alloy 800 SG tubing materials in neutral SG
crevice chemistries with different chloride concentrations
at 40, 150, and 300°C, respectively. The electrochemical
experimental results show that Alloy 800 is immune from
pitting degradation if the chloride ion concentration in the
neutral SG crevice chemistry is equal to 0.01 mol/kg or
lower, except at 150°C. At 150°C, Alloy 800 is still susceptible to pitting corrosion even if the chloride ion concentration in the neutral crevice chemistry is as low as 0.01 mol/
kg. The pitting potential of Alloy 800 SG tubing decreases
with the increase of chloride ion concentration when the
chloride concentration in the neutral crevice chemistry
is higher than 0.01 mol/kg. In neutral crevice chemistries
with a fixed chloride ion concentration, the pitting potential of Alloy 800 decreases as the temperature increases.
4.1. Empirical Equations of Pitting Potential of Alloy 800
SG tubing as a Function of Chloride Ion Concentration
The pitting potential of Alloy 800 in the neutral SG crevice chemistry as a function of chloride ion concentration
could be fitted to empirical equations through a linear
regression followed by an inverse second order regression. These empirical equations can be used to determine the pitting potential of Alloy 800 in neutral crevice chemistry with known chloride ion concentration.
The pitting potential of Alloy 800 SG tubing under secondary side SG neutral crevice chemistry conditions at
300°C can be fitted by the following empirical equations:
The pitting potential of Alloy 800 SG tubing under secondary side SG neutral crevice chemistry conditions at
150°C can be fitted by the following empirical equations:
(8)
Epit = (1244 -7544 [Cl-]) mV
0.01 mol/kg ≥ [Cl-] ≤ 0.10 mol/kg
- 2
= (-193.6
+ 166.1/[Cl
] ) mV≥ [Cl
EpitE=pit(1244
-7544
[Cl ]) mV] – 9.68/[Cl
0.01 mol/kg
[Cl-]] >≤ 0.10 mol/kg
Epit = (-193.6 + 166.1/[Cl-] – 9.68/[Cl-]2) mV [Cl-] > 0.10 mol/kg
(9)
The pitting potential of Alloy 800 SG tubing under secondary side SG neutral crevice chemistry conditions at
40°C can be fitted by the following empirical equations:
Epit = (183 + 141/[Cl-] – 4.87/[Cl-]2)mV [Cl-] > 0.10 mol/kg
(10)
4.2. Empirical Equations of Pitting Potential of Alloy
800 SG tubing as a Function of Temperature
The pitting potential of Alloy 800 in a neutral SG crevice
chemistry with a fixed chloride ion concentration as a function of temperature could be fitted with quadratic equations shown in Equations 11-14:
Epit= (-892 + 6.17 x 1055K-1
) mV
(0.1 M Cl--)
Epit= (-892 + 6.17 x 105 K-1-1) mV
(0.1 M Cl )
Epit= (-892 + 6.17 x 10 55K -1-1) mV
(0.1 M Cl--)=
(-840
+
3.80
x
10
K
)
mV
(0.65
M Cl )
pit
5
-1
EE
=
(-892
+
6.17
x
10
K
)
mV
(0.1
Epitpit= (-840 + 3.80 x 10 K ) mV (0.65MMClCl-)-)
)
mV
(0.65
M
Cl--)
Epit= (-840 + 3.80 x 1055K5-1
-1
= (-964
+ 4.01
x 10
) mV (0.65
(1.0 M
MCl
(-840
+ 3.80
x 10
K5-1K
) -1mV
Cl-)))
EpitE
E=pit
(1.0 MCl
pit= (-964 + 4.01 x 10
5 K
-1 ) mV
Epit= (-964 + 4.01 x 10 55K -1-1) mV (1.0 MCl--)(-937++4.01
3.54xx10
105KK-1))mV
mV (1.0
(4.0MCl
MCl )
EE
==(-964
Epitpit
(4.0 MCl-)-)
pit= (-937 + 3.54 x 10 K ) mV
Epit= (-937 + 3.54 x 1055K-1
)
mV
(4.0
MCl
)
Epit= (-937 + 3.54 x 10 K-1) mV (4.0 MCl-)
4.3. Limitations of the Empirical Equations
(11)
(12)
(13)
(14)
The empirical equations are applicable to Alloy 800 SG
tubing under nuclear SG secondary side crevice or under
deposit conditions for stations using all volatile treatment
(AVT) for SG feed water treatment. The experimental results had good reproducibility and the standard deviations
of the data were less than 50 mV. It is known that certain SG
impurities such as copper, magnesium [12, 13] and thiosulphate [14] will have significant impact on the pitting potential of SG tubing. Therefore, the empirical equations may
not be valid if the SG systems contain the above mentioned
impurities.
Acknowledgements
This work is funded by AECL. The experimental work was
performed by M. Dupuis. The author would like to thank P.
Angell and R.L. Tapping for the management support. M.
Huang is acknowledged for his help on determining the
crevice chemistries for testing and S. Klimas for reviewing
this paper and many helpful suggestions.
AECL NUCLEAR REVIEW
19
aecl Nuclear Review
vol 1, Number 1, june 2012
References
localized corrosion of nuclear grade alloy 800 under steam
generator layup, startup and operating conditions - y. lu
[1] Y.C. Lu, February 2002, “Localized Corrosion and its Prevention for Heat Exchanger Tubing under Steam Generator Crevice Conditions”, Part I. Mechanisms and Electrochemical Approaches for
Establishing Safe Potential/pH Zones, COG Report, COG 01 051
[2] Y. Lu, November 2006, “Minimize Corrosion Degradation of Steam Generator Tube Materials Updated ECP/pH Zone for Alloy 800 SG Tubing ”, Proceedings of the 5th CNS International Steam
Generator Conference, Toronto
[3] S. Liu and Y.C. Lu,“Evaluation of SCC Susceptibility of Alloy 800 under CANDU Steam Generator Secondary –Side Conditions”, CW-127320-CONF-001, Proc. of the 5th CNS International Steam
Generator Conference, Nov. 26-29 2006, Toronto. (2006)
[4] M. Mirzai, C. Maruska, S. Pagan, O. Lepik, G. Ogundele, M. Wright, and G. Kharshafdjian, 1997, “Stress Corrosion Cracking/Corrosion Fatigue/Fatigue in Alloy 600”, Proceedings of the 8th International Symposium on Environmental Degradation of Materials in Nuclear Power Systems Water Reactors, Vol. 1, pp. 11-18, American Nuclear Society Inc., La Grange Park, IL, USA
[5] G.R. Burton and C.W. Turner, 2007, “Rationale for Crevice Chemistry Test Solution Compositions”, COG 06 4057
[6] L. Chi, G. Goszczynski and Y.C. Lu,“Test Plan for Investigating the Low Temperature Degradation of Steam Generator Tubing Materials in the Environments Containing Intermediate Oxidation State
of Sulphur”, 153-33112-COG-009, COG OP-09-4076 (2009)
[7] P.V. Balakrishnan, November 1977, “Effect of Condenser Water in leaking on Steam Generator Chemistry”, Paper Number IWC 77 12, Proceedings of the 38th Annual Meeting of International
Water Conference, Pittsburgh, Pennsylvania, USA
[8] A. Anderko, P. Wang, and M. Rafal, 2002, “Electrolyte Solutions: From Thermodynamic and Transport Property Models to the Simulation of Industrial Processes”, Fluid Phase Equilibria, 194-197,
pp. 123-142
[9] ASTM Designation: “Standard Reference Test Method for Making Potentiostatic and Potentiodynamic Anodic Polarization Measurements”, G5 94 (Reapproved 2011)
[10] R.S. Greeley, W.T. Smith, Jr., R.W. Stoughton, and M.H. Lietzke, 1960, “Electromotive Force Studies in Aqueous Solutions at Elevated Temperatures. I. The Standard Potential of the Silver Silver
Chloride Electrode”, Journal of Physical Chemistry, 64, pp. 652-657
[11] F. King, M.G. Bailey, C.F. Clarke, B.M. Ikeda, C.D. Litke, and S.R. Ryan, 1989, “A High-Temperature, High-Pressure, Silver Silver Chloride Reference Electrode: A User’s Guide”, Atomic Energy of
Canada Ltd. Report, AECL 9890
[12] Y.C. Lu, D. Burns and M.G. Dupuis, 2009, “The Interactive Effects between Steam Generator Impurities on CANDU Steam Generator Tube Corrosion A Final Report”, 153 33112 COG 007 COG
09 4058
[13] Y. C. Lu, 2010, “Effect of Hazardous Impurities on Steam Generator Tube”, ICONE18-30120, Proceedings of the 18th International Conference on Nuclear Engineering (ICONE18), May 17-21,
2010, Xi’an, China.
[14] L. Chi and Y. C. Lu, “Electrochemical Studies of Steam Generator Tube Degradation in The Presence of Thiosulphate”, 153-33112-PLA-003, Proc. of the 15th International Conference on Environmental Degradation of Materials in Nuclear Power Systems-Water Reactors, August 7-11, 2011 Colorado Springs, Colorado
20
FULL ARTICLE
Abstract
Recent cross-section measurements on
gadolinium have raised concerns over the
accuracy of moderator poison reactivity
coefficient calculations. Measurements
have been made at the ZED-2 (Zero Energy
Deuterium) critical facility, Chalk River
Laboratories, AECL, to study the reactivity
effect of gadolinium in the moderator. Since
the neutron capture cross-section of boron
is well known, measurements were also
made with boron to provide calibration
data for measurements with gadolinium.
The measurements have been used to
quantify the bias of the reactivity effect in
full-core simulations of ZED-2 using MCNP, a
neutron transport code used extensively for
simulations of nuclear systems, along with
the ENDF/B-VII.0 cross-section data. The
results showed a bias of −0.41 ± 0.07 mk/
ppm, or −2.1% ± 0.3%, given a reactivity
worth of −20.1 mk/ppm for gadolinium.
Additional simulations also show that the
gadolinium neutron capture cross-section
has been over-corrected, relative to previous
evaluations, in a beta version of ENDF/B
VII.1, which incorporates the Leinweber
data.
nuclear data and the effect of
gadolinium in the moderator
J.C. ChowA*, F.P. AdamsA, D. RoubstovA, R.D. SinghB and M.B. ZellerA
A
B
Atomic Energy of Canada Limited, Chalk River Laboratories, Chalk River, Ontario, Canada, K0J 1J0
Candu Energy Inc., 2280 Speakman Drive, Mississauga, Ontario, Canada, L5K 1B1
Article Info
Keywords: gadolinium cross-section, moderator poison, nuclear data, ZED-2
Article history: Received 24 April 2012, Accepted 9 June 2012, Available online 30 June 2012.
*Corresponding Author: (613) 584-3311 ext. 44437, [email protected]
1. Introduction
With the highest neutron absorption cross-section in the thermal region of all
naturally occurring elements, gadolinium is used in CANDU®1 reactors to suppress
excess reactivity or to shut down the reactor in case of emergency. Recent measurements on total neutron capture cross-section of gadolinium [1] have raised
concerns over the accuracy of moderator poison reactivity coefficient calculations.
It was reported [2] that the use of the Leinweber data helps to improve the prediction of total fission reaction rate as well as the ratio of 238U capture to total fission reaction rate in gadolinium-poisoned fuel (UO2-Gd2O3) in some BWR configurations.
However, in the recent release of ENDF/B-VII.1 evaluated nuclear data library [3],
the Leinweber data were not taken into account. In order to address this issue,
measurements have recently been made at the ZED-2 (Zero Energy Deuterium)
critical facility [4], Chalk River Laboratories, AECL, to study the reactivity effect of
up to 1.5 ppm of gadolinium in the moderator. ZED-2 is a heavy-water-moderated
reactor that operates predominantly in the thermal neutron region. Since the neutron capture cross-section of boron in the thermal region is known to within ±0.1%
[5,6], measurements were also made with up to 6 ppm of boron to provide calibration data for measurements with gadolinium.
The measurements have been used to quantify the reactivity effect of gadolinium in
the moderator in full-core simulations of ZED-2 using MCNP5 [7], a neutron transport code used extensively for simulations of nuclear systems. The objective of the
analysis is to quantify the bias and uncertainty in the effective neutron multiplication factor, keff, in full-core MCNP simulations of ZED-2 as a function of gadolinium
concentration. The MCNP models were executed on a computer cluster recently
installed at the Chalk River Laboratories under the Linux platform. The targeted
nuclear data library for the majority of the analysis is E70CRL [8], developed at
AECL based on ENDF/B VII.0 [9]. However, results based on gadolinium nuclear
data from a beta version of ENDF/B VII.1, which incorporated the results from
Leinweber, as well as the release version of ENDF/B VII.1 [3] will also be presented
for comparison.
2. The ZED-2 Core Configuration
1
The measurements were performed in the ZED-2 critical facility, which is categorized as a tank-type research reactor (see Figure 1). The calandria tank is
an aluminum cylinder of ~3.4 m both in diameter and height, surrounded by a
graphite reflector radially and at the bottom. High purity (>98 wt%) heavy water is used as moderator. Fuel channels can be installed from the top to form
the reactor core. Each fuel channel is installed with five fuel bundles of ~50
cm in length. Criticality is achieved by raising the moderator level in the calandria tank. Major parameters taken for each experimental run include the
critical height, temperature, and purity of the moderator, as well as the temperature of the fuel. Elaborated details about the ZED-2 facility can be found in [4].
CANDU® is a registered trademark of Atomic Energy of Canada Limited (AECL).
AECL NUCLEAR REVIEW
21
aecl Nuclear Review
vol 1, Number 1, june 2012
Figure 1
ZED-2 Reactor Layout.
The configuration of the reactor core for the present measurements is shown in Figure 2. The core consists of 52
fuel channels each of which was installed with five Low
Enriched Uranium (LEU) or Recovered Uranium (RU) fuel
bundles in a 24.5 cm pitch square lattice. The channels are
often filled with water coolant. However, fuel channels for
the present measurements were filled with air to avoid potential complications in the analyses.
In order to maintain a consistent setup for the measurements, the core configuration was left unaltered throughout
the experiment except for the controlled addition of poison
in the moderator. A reference measurement was taken at
the beginning of each series before the addition of poison.
In the main series of measurements, gadolinium was added
to the moderator from nominal values of 0.5 to 1.5 ppm in
increments of 0.5 ppm, and in the reference series with boron from nominal values of 2.0 to 6.0 ppm in increments of
2 ppm.
3. Analysis with MCNP5
The measurements have been analyzed with MCNP5. The
targeted nuclear data library for the analysis is E70CRL [8],
developed at AECL based on ENDF/B VII.0 [9]. The core parameters pertaining to modeling the experimental runs for
the Boron and Gadolinium series are listed in Table 1 and
Table 2, respectively, which were subsequently used, along
with geometric and material data available elsewhere, e.g.,
[4], as input for development of MCNP models.
22
2
3
nuclear data and the effect of gadolinium in the moderator
j.c. chow, f.p. adams, d. roubstov, r.d. singh and m.b. zeller
Figure 2
ZED-2 core configuration for present measurements.
The data from the Boron series was first used to calibrate the system with respective to the poison concentration. Since the neutron capture cross-section of boron in
the thermal region is known to within ±0.1% [6], which is
equivalent to an uncertainty in computed keff of ±0.03 mk2,
the functional dependence of keff to boron concentration is
expected to be negligibly small, on condition that the boron
concentration is properly modeled. In the present measurements, the nominal boron concentrations3 would be biased due to uncertainty in the ZED-2 moderator inventory,
which is known to only ±2%. However, a calibration factor
can be obtained by adjusting the nominal boron concentration value until the keff value matches that of the model
without poison. Hence, the analysis started with developing a reference MCNP model, without poison, according to
Table 1
Boron Series – Core Parameters for MCNP Models
Case #
Boron
Conc.*
[ppm]
Moderator†
Purity
[wt% D2O]
Channel
Temperature
[C]
Moderator
Temperature
[C]
Critical¶
Height
[cm]
B1
0.0
98.743
21.50
21.55
131.650
B2
2.0
98.742
21.30
22.05
139.420
B3
4.0
98.740
21.63
22.19
148.248
B4
6.0
98.738
22.15
22.53
158.442
* These are nominal values based on the mass of the poison added relative
to the known inventory of moderator in the system.
†
Standard deviation in moderator purity = 0.005 wt% D2O.
¶
Standard deviation in critically height = 0.2 cm.
This value was obtained by a sensitivity study of a model with 6 ppm of boron in the moderator.
The nominal concentration was determined as the ratio of the mass of poison added to the moderator to the recorded moderator inventory.
aecl Nuclear Review
vol 1, Number 1, june 2012
nuclear data and the effect of gadolinium in the moderator
j.c. chow, f.p. adams, d. roubstov, r.d. singh and m.b. zeller
Table 2
Gadolinium Series – Core Parameters for MCNP Models
Case #
Gd
Conc.*
[ppm]
Moderator†
Purity
[wt% D2O]
Channel
Temperature
[C]
Moderator
Temperature
[C]
Critical¶
Height
[cm]
G1
0.0
98.748
21.80
21.70
131.585
G2
0.5
98.748
21.75
21.79
138.248
G3
1.0
98.744
21.65
21.89
145.632
G4
1.5
98.739
21.45
22.51
153.926
* See footnotes in Table 1.
the configuration described in Section 2, followed by a case
with the maximum nominal value of 6.0 ppm of boron in the
moderator. Then, the boron content was adjusted until the
keff value matched that of the case without poison to within
±0.1 mk, the statistical uncertainty inherent in the MCNPcomputed keff values obtained with 100 million neutron histories. The adjustment in boron concentration to achieve
the above condition was found to be −1.2%. The same adjustment was then applied to the other cases with nominal
boron concentrations of 2.0 and 4.0 ppm. The resulting keff
values with the adjusted concentrations are listed in Table
3 and plotted in Figure 3 as a function of boron concentration. The relatively small slope (0.003 mk/ppm) in the line
in Figure 3 indicates that the functional dependence of the
keff values on the boron concentration can be minimized,
within uncertainty, with proper adjustments to the nominal
values. Thus, the nominal boron concentrations are judged
to have been overestimated due to uncertainty in the moderator inventory, and a correction factor of −1.2%, henceforth referred to as the boron calibration factor, should be
applied to the nominal boron concentration values.
Since the Boron and Gadolinium series were performed
consecutively with the same moderator inventory and the
same protocols were used for measuring and adding the
poisons to the moderator, the boron calibration factor can
also be applied to the Gadolinium series of measurements.
The results of keff values obtained by applying the boron
calibration factor, i.e., multiplying the nominal gadolinium
concentration values by the factor 0.988, are shown in Table 4. A discussion on the results follows the assessment of
uncertainties in Section IV below.
Table 3
Boron Series − MCNP keff Values with Adjusted Concentrations
B1
Adjusted Boron
Conc. [ppm]
0.000
0.99785
B2
1.976
0.99788
B3
3.952
0.99800
B4
5.928
0.99779
Case
keff
Table 4
Gadolinium Series − MCNP keff Values with Boron-Calibrated Concentrations
G1
Boron-Calibrated
Gd Conc [ppm]
0.000
0.99788
G2
0.494
0.99766
G3
0.988
0.99759
G4
1.482
0.99713
Case
keff
0.9990
4. Assessment of Uncertainties
y = ‐0.000003x + 0.99790
0.9985
keff
0.9980
0.9975
0.9970
N.B.: Error bars = ±2 × stat. unc.
0.9965
0
2
4
Boron Conc. [ppm]
6
Figure 3
Boron series − MCNP keff values with adjusted Boron concentration.
5
The total uncertainty in keff (σkeff) obtained in an MCNP simulation of a ZED-2 full-core benchmark model, using fuel
channels4 similar to those in the present experiment, was
assessed by Atfield [4] to be about ±3 mk, consistent with
the biases in keff values observed in Table 3 and Table 4. In
the following assessment, only terms that contribute to the
uncertainty in keff bias due to the presence of gadolinium in
the moderator are included.
The uncertainty in the computation of keff bias relevant to
the present analysis is determined as follows:
 k   2stat   2 expt   2 cal   2 Gd
eff
The uncertainties in the dimensions of the fuel channels were found in reference [4] to be the largest contributors to the keff uncertainty.
AECL NUCLEAR REVIEW
(1)
23
nuclear data and the effect of gadolinium in the moderator
j.c. chow, f.p. adams, d. roubstov, r.d. singh and m.b. zeller
where σstat is the statistical uncertainty intrinsic to the
MCNP simulation, σexpt is the experimental uncertainty, σcal
is due to using the boron calibration factor to calibrate the
gadolinium concentration, and σGd is due to uncertainties in
the gadolinium neutron capture cross-section.
The statistical uncertainty arises from the Monte Carlo nature of simulations using MCNP. All the MCNP models have
been executed with 100 million active neutron histories
yielding a statistical uncertainty, σstat, of ±0.06 mk.
Assuming no correlation among the contributing terms, the
experimental uncertainty, σexpt, can be expressed as:
 expt  
2
hc

2
pur
(2)
where σhc and σpur are uncertainties due to the critical
height and purity of the moderator. Based on a sensitivity
analysis by direct perturbation calculations similar to those
performed in reference [4], σhc and σpur have been assessed
as ±0.036 mk and ±0.065 mk (see footnote in Table 1), respectively, yielding an experimental uncertainty, σexpt, of
±0.07 mk.
The uncertainty in the boron calibration factor is assessed
as contributed by two terms; one is due to the statistical
uncertainty6 of ±0.06 mk intrinsic to the MCNP simulation;
the other is due to uncertainty in the total boron neutron
capture cross-section of ±0.1% in the thermal region [5,6],
which is equivalent to an uncertainty in keff of ±0.03 mk for
a model with 6 ppm of boron (with a total reactivity worth
of 33.6 mk, obtained by adding 6 ppm of boron to the reference ZED-2 model without boron and comparing the keff
values). Adding the two terms in quadrature yields an uncertainty of ±0.07 mk, or ±0.2% of the total boron reactivity
worth of 33.6 mk. Therefore, the boron calibration factor
is assessed as −1.2% ± 0.2%. By adding one ppm of gadolinium to the base model without gadolinium and comparing the keff values, the reactivity worth of gadolinium in
the moderator in ZED-2 has been determined as −20.1 mk/
ppm. Therefore, σcal, the uncertainty in keff due to the use of
the boron calibration factor, is assessed as ±0.04 mk/ppm
(= 20.1 mk/ppm × 0.2%).
24
The uncertainty attributable to the gadolinium neutron
capture cross-section has been investigated from various sources. Figure 4 shows the uncertainties around
the thermal region for 157Gd from three sources. Note
that only 157Gd, which contributes ~80% of the total neutron capture cross-section, is included in this assessment.
While Mughabghab [10] assigns a single value of ±0.32%
at 0.0253 eV, ORNL [11] assigns values ranging from 0.3%
to ~3%, and NNDC [12] assigns a value of ±4% for the entire thermal region. For the purpose of the present analysis, a value of ±2% is adopted. A sensitivity analysis of the
three models containing gadolinium in the moderator has
6
shown that an uncertainty of ±2% in the gadolinium neutron capture cross section is equivalent to a keff uncertainty,
σGd, of 0.50 mk/ppm. All the uncertainty components contained in Equation (1) and obtained according to the above
assessments are listed in Table 5. It should be reiterated
that only terms that contribute to the uncertainty in keff bias
due to the presence of gadolinium in the moderator are included in the above assessments.
5
Capture Cross‐Section Uncertainty [%] [%]
aecl Nuclear Review
vol 1, Number 1, june 2012
4
3
2
ENDF/B‐VII.1 (NNDC)
1
ENDF‐B/VII.0 (ORNL)
0
0.001
0.01
0.1
1
10
Neutron Energy [eV]
Figure 4
157
Gd Neutron Capture Cross-Section Uncertainty
Table 5
Sources of keff Uncertainties
Symbol
Contrib. to
σkeff
MCNP Statistical Uncertainty
σstat
±0.06 mk
Experimental (moderator height/purity)
σexpt
±0.07 mk
Boron calibration
σcal
±0.04 mk/ppm
157
σGd
±0.5 mk/ppm
Sources
Gd Neutron Capture Cross-section
5. Discussion of Results and Conclusions
The results of keff values obtained with boron-calibrated
gadolinium concentrations are listed in Table 4. Note that
the bias in keff values of ~−2 mk, attributable mainly to uncertainties in the dimensions of the fuel channel and fuel
enrichment [4], would not affect the present results, which
are based on relative measurements/calculations. A plot
of keff values verses gadolinium concentrations is shown in
Figure 5, in which the uncertainties are assigned according
to Table 5. According to a linear regression analysis of the
data points weighted with their respective variances, the
bias in keff value in MCNP simulations of the ZED-2 full-core
with gadolinium in the moderator is assessed as −0.41 ±
0.07 mk/ppm, or −2.1% ± 0.3%, for the target nuclear data
library ENDF/B-VII.0, given a reactivity worth of −20.1 mk/
ppm for gadolinium in ZED-2.
Note that this statistical uncertainty arises from simulation of a model containing boron while σstat in Equation (1) arises from a model containing gadolinium.
aecl Nuclear Review
vol 1, Number 1, june 2012
While the above results were obtained with the E70CRL nuclear data library [8] based on ENDF/B VII.0 [9], additional
MCNP simulations of the same models but incorporating
the 157Gd cross section data from the release version, as
well as a beta version, of the ENDF/B VII.1 data library have
also been performed. In the beta version, the measurements from [1] were taken into account, in which the scattering width (  n) of the first resonance at 0.0314 eV was
decreased by 7% from previous evaluations. The results
of the additional simulations with 157Gd cross section data
from the ENDF/B VII.1 library are plotted in Figure 6 along
with the results obtained with ENDF/B VII.0 [9]. While the
results based on the release version of ENDF/B VII.1 are
identical to those based on ENDF/B VII.0 within statistical
uncertainties, those from the beta version of ENDF/B VII.1,
which incorporates the data from Leinweber [1], deviate
significantly from those based on the ENDF/B VII.0 data. In
essence, Figure 6 shows that the gadolinium neutron capture cross section is overestimated in ENDF/B VII.0 [9] data
and underestimated in the beta version of ENDF/B VII.1
which incorporates the Leinweber [1] data, indicating that
the decrease of 7% might have been overcompensated.
nuclear data and the effect of gadolinium in the moderator
j.c. chow, f.p. adams, d. roubstov, r.d. singh and m.b. zeller
Figure 6
Comparison with ENDF/B-VII.1 data.
Based on the above results, it is concluded that the reactivity effect of gadolinium in the moderator can be calculated
with MCNP5, along with the ENDF/B-VII.0 or ENDF/B-VII.1
nuclear data libraries, with a bias that is consistent with the
nominal ±2% uncertainty of the nuclear data. The incorporation of the Leinweber data in the ENDF/B-VII.1 beta
evaluation changed the sign of the apparent bias in the calculations, but the overall accuracy was not improved. The
results provide confidence in the use of the ENDF/B-VII.0
or ENDF/B-VII.1 nuclear data evaluations in calculations of
reactivity effect of gadolinium in the moderator.
6. Acknowledgements
Figure 5
Gadolinium series −MCNP keff values with Boron-calibrated
concentrations.
We thank J. Atfield for help with the pre-experiment simulations, G. Cully, K. Thompson, and D. Grice for conducting
the experiments, and L. Blomeley for compiling the experimental data which formed the basis of this work. We also
thank the EC6 Group of Candu Energy Inc. for funding the
experiment.
References
[1] G. Leinweber et al., 2006, “Neutron Capture and Total Cross-Section Measurements and Resonance Parameters of Gadolinium”, Nuclear Science and Engineering, 154, pp. 261-279
[2] G. Perret et al., 2009, “Impact of New Gadolinium Cross Sections on Reaction Rate Distributions in 10 × 10 BWR Assemblies”, Nuclear Science and Engineering, 163, pp. 17 25
[3] M.B. Chadwick et al., 2011, “ENDF/B-VII.1 Nuclear Data for Science and Technology: Cross Sections, Covariances, Fission Product Yields and Decay Data”, Nuclear Data Sheets, 112, pp. 28872996
[4] J. E. Atfield, March 2011, “28-Element Natural UO2 Fuel Assemblies in ZED-2 (ZED2 HWR EXP 001)”, in International Handbook of Evaluated Reactor Physics Benchmark Experiments (IRPhEP),
OECD – NEA/NSC/DOC (2006)
[5] International Atomic Energy Agency, Vienna, 2007, “International Evaluation of Neutron Cross-Section Standards”, STI/PUB/1291
[6] A.D. Carlson et al., 2009, “International Evaluation of Neutron Cross Section Standards”, Nuclear Data Sheets, 110, pp. 3215–3324
[7] X-5 Monte Carlo Team, April 2003, “MCNP - A General Monte Carlo N-Particle Transport Code, Version 5. Volume 1: Overview and Theory”, LANL report LA-UR-03-1987 (Revised 10/3/2005)
[8] D. Altiparmakov, May 2010, “ENDF/B-VII.0 versus ENDF/B-VI.8 in CANDU® Calculations”, Proceedings of PHYSOR 2010 – Advances in Reactor Physics to Power the Nuclear Renaissance, Pittsburgh, Pennsylvania, USA, on CD ROM, American Nuclear Society, LaGrange Park, IL, USA
[9] M.B. Chadwick et al., 2006, “ENDF/B-VII.0: Next Generation Evaluated Nuclear Data Library for Nuclear Science and Technology”, Nuclear Data Sheets, 107, pp. 2931-3060
[10] S.F. Mughabghab, “Atlas of Neutron Resonances – Resonance Parameters and Thermal Cross Sections. Z=1-100”, 5th Ed. Elsevier Science
[11] Oak Ridge National Laboratory, 2009, “SCALE: A Modular Code System for Performing Standardized Computer Analyses for Licensing Evaluation”, ORNL/TM 2005/39, Version 6
[12] NNDC, “National Nuclear Data Center, 2011, Evaluated Nuclear Data File (ENDF) Retrieval & Plotting”, http://www.nndc.bnl.gov/sigma/index.jsp, Brookhaven National Laboratory, ENDF/BVII.1
AECL NUCLEAR REVIEW
25
26
FULL ARTICLE
Abstract
The positions of the components of a reactor
can change over time, due to radiation
damage, sagging, etc. Thus, it is important
to determine their positions. To satisfy this
requirement of the staff at the Point Lepreau
Generating Station, a method to determine
the positions of reactor components has been
developed and demonstrated. This method
combines the use of dose rate measurements
measurements of the high dose
rate profiles inside a shutdown
candu® reactor
C. JewettA*, J. ChowA, D. ComeauB, G. JonkmansA, B. SmithA, B. SurA,
D. TaylorB, S. YueA
A
B
Atomic Energy of Canada Limited, Chalk River Laboratories, Chalk River, Ontario, Canada, K0J 1J0
Point Lepreau Generating Station, Point Lepreau, NB, Canada, E5J 2S6
Article Info
Keywords: gamma-ray imaging; radiation measurement; CANDU® reactors; MCNP5 simulations; gamma-ray detectors
* Corresponding author: (613) 584-3311, [email protected]
and Monte Carlo simulations. It first involves
measuring the high γ-ray dose rates as a
function of position within a reactor. Then it
entails comparing these measurements with
Monte Carlo simulations. In order to perform
such measurements, a silicon diode detector
and a scan drive system have been developed.
In 2009, measurements of the γ-ray dose
rate profile of the shut down Point Lepreau
Generating Station reactor were conducted.
By comparing the locations of the local
peaks in the dose rate data, it was possible
to determine the distances between the
steel reactor components. The measured
data were then compared with Monte Carlo
simulations to determine how precisely one
could locate the positions of the adjuster rods.
Using this technique, it was found that the
retracted adjuster rods were 440 ± 60 mm
below their designed positions.
1. Introduction
Since it is possible for the components of a reactor to shift over time, it is necessary
to be able to determine their positions accurately. For example, the positions of the
components of a reactor can have an impact on the neutron flux shape within the core
of the reactor. Thus, the primary objective of this work was to accurately determine
the positions of Adjuster Rods 15 and 16 (See Figure 2). It was also necessary to obtain accurate information regarding the radiation field distribution inside a reactor
for a number of reasons. One of these reasons is the establishment of safe work plans,
based on the doses workers would receive. Another reason is that it is important
to know the highest levels of radiation that the in-core instrumentation will experience. If the radiation field is too intense for a detector to perform properly, then it
must be replaced by one that will, so that safe operation of the reactor is guaranteed.
A method to create a 1-dimensional radiation profile of the interior of the PLGS
reactor has been proposed and implemented. By profiling the interior of the reactor, it is possible to determine the positions of features, such as the adjuster rods,
of the reactor. It is important to know the positions of these features, since they
have an impact on the operation of the reactor. While the measurements described
in this paper were performed on a shutdown reactor that was undergoing refurbishment, the plan is to perform these measurements again on the same reactor
shortly before it is started up again. Knowing the positions of the adjuster rods
will aid in the understanding of any neutron flux shape tilts that may occur. The
measurements on the shutdown reactor were performed to demonstrate the viability of this method. This method employs a combination of measurement of
the γ-ray dose rate as a function of position within one of the access pipes in the
reactor core (called Startup Instrumentation (SUI) tubes), and simulations of
that dose rate in MCNP5, a Monte Carlo radiation transport program [1]. Monte
Carlo simulations were performed to aid in the understanding of the measured
data, and assist in more accurately determining the positions of the adjuster rods.
Since the geometry of the reactor is rather complex, developing an analytic solution to the γ-ray emission and transport problem is extremely difficult. In addition to this, the anisotropic nature of γ-ray scattering further complicates the γ-ray
transport problem. Thus, Monte Carlo was chosen as the modeling technique.
This paper begins by describing the γ-ray dose rate detection system. It then describes the measurements performed at the PLGS reactor while it was being refurbished. Finally, this paper discusses the MCNP5 simulations, and compares the
simulation results with the measured data to draw conclusions about the positions of the components of the reactor with respect to their assumed positions.
AECL NUCLEAR REVIEW
27
aecl Nuclear Review
vol 1, Number 1, june 2012
measurements of the high dose rate profiles inside a shutdown candu reactor
c. jewett, j. chow, d. comeau, g. jonkmans, b. smith, b. sur, d. taylor and s. yue
2. The γ-ray Detector System
A small, inexpensive gamma-radiation detector, based
on a commercially available Silicon (Si) p-n junction diode, has been developed. This detector can accurately
measure γ-radiation fields over a wide dynamic range
from 0.2 to 600 Gy/hr [2] [3] [4] The current of the Si diode provides a measure of the γ-ray dose received by the
diode, since the current produced by the diode is proportional to the ionization energy deposited in it [5].
The detector consists of a disc-shaped diode (1 mm in diameter and 0.3 mm thick), encapsulated within a steel jacket. The very small, slender shape of the detector enables it
to meet the requirement that it make point-like measurements of the dose rate. Tests at the Chalk River Laboratories
of Atomic Energy of Canada Limited (AECL) demonstrated
the ability of the diode sensor to perform well in radiation
fields as high as 600 Gy/hr. These tests also demonstrated
the ability of the detector to survive a total absorbed dose
of 10 kGy with only a 5% decrease in dose rate response [3].
Calibrating the detector requires knowledge of the relationship between the current and the γ-ray energy absorbed
(dose) by the diode per unit time. The current was measured as a function of γ-ray dose rate by exposing the detector to a nominally 3.7×1011 Bq, point-like Co-60 source.
The dose rate was varied from 0.2 Gy/hr to 2 Gy/hr by
changing the distance between the detector and the Co60 source. Figure 1 shows a plot of the measured detector
current versus dose rate. The dose rate response function
of the detector is linear, in accordance with expectations
[5]. The current-to-dose rate calibration factor was 28.3
pA/(Gy/hr) before the detector was exposed to the reactor. While it was in the PLGS reactor, the diode received a
maximum dose rate of 1079 Gy/h. After the detector was
exposed to the reactor, the current it emitted was re-measured as a function of dose rate by using another calibrated
Co-60 source. The results of these measurements, which
appear in Fig. 1, revealed a decrease in the current-to-dose
rate calibration factor to 27.5 pA/(Gy/hr).
An automated scan drive system was built to raise and lower the diode detector within the start up instrumentation
(SUI) tube. The scan drive system consisted of four parts:
a take-up reel, a stepper motor, a modified eddy current
pusher and a rotary encoder. The Mineral Insulated (MI)
cable was wrapped around the take-up reel. The stepper
motor, which the data acquisition (DAQ) computer controlled via a motor controller, turned both the take-up reel
and the four rubber wheels of the eddy current pusher. The
rubber wheels of the eddy current pusher gripped the MI
cable, and pushed or pulled it, depending on the motions
of the stepper motor to which it and the take-up reel were
geared. Finally, the rotary encoder, which touched the surface of the take-up reel, measured the position of the detector based on the distance take-up reel rotated. This system
could position the detector to within 1 mm for every metre
the diode travelled [4].
A Keithley 6487 picoammeter, with a dynamic range of 0.1
nA to 30.1 nA, was used to read the current from the detector. The error on the measured currents ranged from 0.001
nA to 0.46 nA [6]. After reading the detector’s current, the
Keithley sent its value to the Data Acquisition (DAQ) computer. In addition to the current signal, the DAQ computer
also recorded the time of each measurement and the position of the detector.
3. Measurements of the γ-radiation Profile of the Reactor
Figure 1
Plots of the diode detector current signal versus dose rate
calibration data for the silicon diode detector. The offset between the y-intercepts of the two lines is due to the instrument offset of the Keithley 6487, and thus has no physical
meaning.
28
A radiation profile measurement of the PLGS reactor was
made to determine the positions of the withdrawn adjuster
rods. The radiation dose rate profile of the shutdown reactor depended on both the operational history of the reactor and the amount of time that the reactor had been shutdown when the dose rate profile measurements were made.
Thus, the positions of the adjuster rods and other features
of the reactor were determined from the relative values of
the measured dose rate. Nevertheless, the diode was calibrated to provide absolute values of the dose rates in the
reactor for safe work planning.
The most basic parts of a CANDU® reactor are: the calandria, the core, the shielding surrounding the calandria,
and the control and diagnostic instruments of the reactor.
aecl Nuclear Review
vol 1, Number 1, june 2012
measurements of the high dose rate profiles inside a shutdown candu reactor
c. jewett, j. chow, d. comeau, g. jonkmans, b. smith, b. sur, d. taylor and s. yue
The calandria, which is made of steel, holds the core. The
core holds the fuel channels, and is filled with heavy water.
The heavy water acts as a moderator, which slows downs
the fast fission neutrons until they have thermal energies.
During the measurements that this paper describes, the
reactor had been shut down, and its core had been emptied of all of the fuel channels and heavy water. Thus, while
the radiation profile measurements were being conducted
in 2009, the calandria contained only air and a few access
pipes. The access pipes enable the control and diagnostic
instruments of the reactor to enter the core. The shielding that surrounds the calandria consists of light water
(see Figure 2), surrounded by a 1220 mm thick concrete
enclosure. Examples of diagnostic instruments that enter
the access ports are neutron flux detectors, which provide
the operators with neutron flux information, while the reactor is running. The adjuster rods are an example of control
devices that enter the access ports. The adjuster rods and
other control and diagnostic devices had been completely
withdrawn from the core into the light water shielding during the 2009 radiation profile measurements [7].
The reactivity mechanisms (RM) deck of the PLGS reactor
provided the platform for the γ-ray dose rate profile measurements (see Figure 2). The γ-ray scan equipment and DAQ
were placed upon the RM deck, while the scan equipment
moved the detector up and down the in-core start-up instrumentation (SUI) tube that was ported to viewport 2 (VP2).
The starting position of the detector was about 0.5 metres above the centreline of the core of the reactor (see
Figure 2). The scan drive system took measurements
from this position to a point 9970 mm above it in 10 mm
increments. At each position, 30 current measurements
were performed to reduce the noise in the current data.
While the reactor was running, the components of the reactor that were in the core absorbed the neutrons generated
by fission reactions in the fuel. These reactor components
are made of steel, aluminum and zirconium alloys. While
steel is a strong neutron absorber, due to the large (n,γ) thermal cross sections of the isotopes of which it is composed,
zirconium alloys and aluminum are rather weak neutron
absorbers [8]. After absorbing the neutrons generated in
the fuel, the reactor components became activated via (n,γ)
reactions. These activation products range from Fe-55 to
Ni-59 to Si-28, etc. Most of these activation products either
have very short or very long half-lives, and thus, had either
completely decayed away by the time of the measurement,
or exhibited extremely low decay rates. The two activation
products whose half lives are intermediate in duration are
Fe-55 (999 d) and Co-60 (1925 d). Fe-55 emits mostly 6
keV X-rays, while Co-60, on the other hand, emits 1173 and
1332 keV γ-rays. Thus, the decay of Co-60 was the dominant source of radiation in these measurements. The Co-59
concentration in the zirconium alloys is typically 20 ppm or
less [8]. The Co-59 concentration in the steel used in a CANDU® reactor is up to 700 ppm for the calandria shell and
nozzles, and is estimated to be 2000 ppm for the adjuster
rods. Therefore, the dominant sources of Co-60 γ-rays were
the irradiated steel components of the reactor. These include the calandria walls and the adjuster rods. Brand new
SUI tubes had been installed just before the measurements,
and hence did not contain any Co-60. In the case of the radiation profile measurements, the calandria and the two
adjuster rods adjacent to the VP2 SUI tube were the dominant sources of γ-rays. Figure 2 is a diagram of the reactor
components that were germane to these measurements.
4. Experimental Analysis and Results
Figure 2
A diagram of the reactor features relevant to the PLGS dose
rate profile measurements. The diagram includes the RM
deck, the outer calandria wall, VP2 SUI guide tube and Adjuster Rods 15 and 16. In this diagram, the adjuster rods
are in their fully retracted states.
To minimize the impact of ambient electronic noise at the
reactor facility, thirty current readings were taken at each
detector position (10 mm increments). The statistical mode
was then taken as the detector current for each set of five
consecutive positions covering a 50 mm range. The mode
rather than the mean was chosen, since the mean values
AECL NUCLEAR REVIEW
29
measurements of the high dose rate profiles inside a shutdown candu reactor
c. jewett, j. chow, d. comeau, g. jonkmans, b. smith, b. sur, d. taylor and s. yue
aecl Nuclear Review
vol 1, Number 1, june 2012
were very noisy. The full width at half maximum (FWHM)
of the current for each of these positions was also determined. The upper and lower errors on each mode current
were then obtained from the differences between the mode
current and the upper and lower limits of the FWHM. The
modes were determined over 50 mm intervals, since these
were the smallest intervals for which the FWHM limits were
reasonably symmetrical and not oversized. The noise in the
current data thus had a limiting effect on the accuracy of the
position data. The error on the position, σz, is thus given by
 
2
z
2
scan

2
range
(1)
The error on the ability of the scan drive to position the
detector, σscan, is equal to Δz × (1 mm/m), where Δz is
the distance in metres the detector travelled to reach
the position z. The quantity σrange is taken as half the 50
mm interval range. Based on Equation 1, the accuracy
on the position varied from 25 mm for measurements
near the bottom of the SUI tube to 27 mm for measurements near its top. The detector positions were taken
as the central positions over each 50 mm mode interval.
The current mode values were converted to dose rates via
linear current-to-dose rate conversions obtained from the
calibrations described in Section 2. These linear conversion
formulae consisted of the following: D1 = m1I and D2 = m2I.
D1 is the linear fit obtained from the calibration data taken
with the diode sensor before it was exposed to the reactor,
and D2 is the linear fit obtained after the diode was exposed
to the reactor. The slopes of these two lines are m1 = 35.3731
Gy/hr/nA and m2 = 36.3771 Gy/hr/nA. The small increase
in the slope of the calibration line is due to a gradual degradation of the ability of the diode to produce current as it
was exposed to the radiation in the reactor. Thus, to obtain
a better estimate of the dose rate, D, of the reactor from the
measured currents and the two sets of calibration data, it
was calculated from the error-weighted average of D1 and D2:
 D
D
D   21  22
 D  D
2
 1
 2
 D


noise in the current readings is ΔFWHM. Its value for a
given position was calculated by taking the larger of
the two differences between the mode current and
the upper and lower FWHM limits. Thus, as one can
see from Equation 4, the error on the dose rate, D, depends upon the measured current, I, and the two calibration factors, m1 and m2. The resulting plot of measured
dose rates versus detector positions appears in Fig. 3.
1

2
D

1
1
 2 2
2 2
I   m1  I I  m2  m22 I2
2
2
m1
(4) [9]
The horizontal axis of Fig. 3 was divided into three sections, named A, B and C. The largest measured dose rate of
1079±36 Gy/h occurred at a detector position of 17013±2.5
mm. The calandria (see Fig. 2), was the dominant source of
Co-60 γ-rays. The main calandria shell alone contains about
2.7×106 cm3 of steel. In contrast, the two adjuster rods on
either side of VP2 contain only about 1800 cm3 of steel.
Since the calandria is the dominant source of γ-rays, and the
calandria edge is the part of the calandria that is closest to
the VP2 SUI tube, the inner edge of the calandria is expected
(2)
The errors, σD, on the dose rates were calculated via the following formula:
1

30
2
D

1

2
D1

1
 D2 1
(3)
The errors on D, D1, D2 and the measured current are σD, σD1,
σD2 and σI [9] . The errors on D1 and D2 were simply calculated
from σD12 = I2σm12+m12σI2 and σD22 = I2σm22+m22σI2. The errors
on the current modes can be found from σI2 =σKeithley2+ΔFWHM2.
The variable σKeithley represents the error due to the uncertainty in the Keithley 6487 reading. The error due to the
Figure 3
Top: A plot of the γ-ray dose rate profile measured within
the VP2 SUI tube. Bottom: A plot of the measured γ-ray dose
rate profile with the contribution of the calandria subtracted.
aecl Nuclear Review
vol 1, Number 1, june 2012
measurements of the high dose rate profiles inside a shutdown candu reactor
c. jewett, j. chow, d. comeau, g. jonkmans, b. smith, b. sur, d. taylor and s. yue
to be the position at which the dose rate was the largest.
Therefore, in the analysis, it was assumed that the 17013
mm sensor position is the location at which the VP2 SUI tube
meets the inner edge of the calandria wall. This 17013 mm
peak position also marks the upper boundary of section A,
which covers the calandria from its interior to its edge. The
position at 17013 mm was used as a reference point to help
in the determination of the locations of Adjuster Rods 15
and 16, which sit on either side of the VP2 SUI tube. Section
B represents the locations between the edge of the calandria
wall and the point at 16095 mm, which is the local minimum
in the dose rate due to the calandria and the adjuster rods.
Section C covers the detector positions from 16095 mm to
the point (12719 mm) at which the γ-ray field due to the
adjuster rods first drops to 0.01 Gy/hr. Thus, the dose
rate in Section C was assumed to be due to the adjuster
rods alone, since they sit on either side of the VP2 SUI
tube. The longer adjuster rod, Adjuster Rod 16, is 3429
mm long, and the shorter adjuster rod, Adjuster Rod 15,
is 1143 mm long. All of the adjuster rods had been fully
withdrawn from the core during these measurements.
After converting the measured detector currents to dose
rates, and determining the location of the calandria edge
in the measured dose rate profile, the locations of Adjuster
Rods 15 and 16 were extracted from the data. The first step
in this process involved fitting a combined exponential-inverse square curve to the calandria dose rate data in Section
B of the Fig. 3 plot. The equation obtained with this fit is
D


exp  3.49  10 3 z
2.86  10  8 z 2  1.18  10 6 z  8.25  10  4
(5)
In this formula, D is the fitted dose rate and Δz is the vertical distance above 17013 mm, the location of the calandria
dose rate peak. The numerator of Equation 5 represents
the exponential absorption of the γ-rays as they propagate
through the water shielding above the calandria shell. The
denominator, which is a quadratic in Δz, represents the
inverse square decrease in the intensity of the γ-ray field
with distance from the calandria shell. The fitted curve
was then subtracted from the dose rate data to obtain the
contributions solely due to the adjuster rods. In this fashion, the adjuster rod dose rate plot in Fig. 3 was obtained.
The centroid of the dose rate peak due to the adjuster rods
appears at about 14868 mm. Assuming that the centres of
Adjuster Rods 15 and 16 were aligned with the centroid of the
dose rate peak, this suggests that the centres of the adjuster
rod were 2145 mm above the calandria edge. The expected
position of the centres of the retracted adjuster rods, based
on the reactor commissioning data, was 2565.5 mm above
the calandria edge (see Fig. 2). This suggests the possibility
that the adjuster rods were out of position by 400 to 500 mm
when in their fully retracted state. This deviation from the
expected value of the positions of the adjuster rods is well
beyond the 70 mm uncertainty which applies to the fully inserted position. In order to better understand the dose rate
profile measurements, and more accurately determine the
positions of the adjuster rods, simulations of the γ-ray dose
rate profile of the PLGS reactor were performed in MCNP5.
5. Simulations with MCNP5
The following components of the shutdown PLGS reactor
were simulated in MCNP5: the concrete shielding, the light
water inside the concrete shielding, the calandria walls, the
air inside the calandria, the VP2 SUI tube, a column of Si
tally cells within the VP2 SUI tube, and Adjuster Rods 15
and 16.
In order to isolate the contributions of each activated steel
component to the total γ-ray dose rate, one MCNP5 input
file was created for each steel part. Accurately modeling the
γ-ray sources due to these reactor components required the
determination of the amount of Co-60 that had been produced in them while the reactor was running. While the
reactor was running, Co-60 was produced by the capture
of neutrons by Co-59, and decayed with a half-life, λCo-60, of
1925 days. Equation 6 shows this production and decay
relationship. The product,  nσCo-59, is average neutron flux
times neutron capture cross section. This product gives the
total 59Co(n,γ)60Co reaction rate. It is given by Equation 7,
which is the sum or integral of the reaction rates at all of the
neutron energies. Solving Equation 6 for NCo-60 yields Equation 8, which gives the number of Co-60 nuclei as a function
of time.
dN Co  60 ,i
dt
  n  Co  59   Co  60
 n Co 60    n E , r  n , E dE
N Co 60 ,i t  

V i  Co  59   t
e
 e Co 60 t
 Co 60  
(6)
(7)

(8)
In the above formulae, NCo-60,i(r) is the amount of Co-60
as a function of position in reactor component i. Vi is the
volume of reactor component i and ρCo-59 is the density of
Co-59. The value of t is the total reactor run time. Finally,
 (E,r) is the neutron flux as a function of energy and position, and σn,γ(E) is the neutron capture cross section for
Co-59. The values of  σ for the steel components were then
calculated by running core criticality calculations of a generic CANDU® core model in MCNP5. Having obtained the
AECL NUCLEAR REVIEW
31
aecl Nuclear Review
vol 1, Number 1, june 2012
measurements of the high dose rate profiles inside a shutdown candu reactor
c. jewett, j. chow, d. comeau, g. jonkmans, b. smith, b. sur, d. taylor and s. yue
simulated Co-60 density distributions within the calandria
walls, nozzles and adjuster rods, these distributions were
then used to simulate the γ-ray sources that were present
in the PLGS reactor when the measurements were made.
6. Analysis of the Data Obtained via MCNP5 Simulations
Each γ-ray simulation provided the contribution of one of the
steel components to the dose in each silicon tally cell in the
SUI tube. MCNP5 provided these tally doses in units of MeV/
g/γ-ray. Thus, in order to determine the contributions of the
steel components to the total dose rate, it was necessary to
determine the rate at which each component emitted γ-rays,
which can be determined from the criticality calculation tallies, Equation 8, and the rate at which the reactor produced
fission neutrons as a function of time and position [10].
The fission neutron production rate is given by Equation 9
P
dn
 th 
dt
E rec
(9)
the measured dose rate data. These cobalt impurity concentrations are all reasonable for stainless-steel alloys.
According to PLGS staff, the cobalt concentration in the
stainless steel of the adjuster rods was about 2000 pp. The
cobalt concentration in the calandria shell was given as 700
ppm. Typical cobalt impurity concentrations in stainless
steel range from 10 ppm to 700 ppm. The largest reasonable cobalt concentration in a stainless-steel alloy is about
2000 ppm. The errors on the cobalt impurity levels in the
calandria shell pieces were estimated by calculating the
dose rates with all of the initial cobalt concentrations set to
10 ppm and then to 700 ppm. The upper and lower error
bars on the simulated calandria dose rate data were then
obtained by calculating the differences between the dose
rates obtained with the lower and upper values of the cobalt concentration and the values used to obtain Figure 4
and Figure 5. The errors on the cobalt impurity levels in the
adjuster rods were estimated by taking half the difference
between 700 ppm and 10 ppm.
The average thermal power of the PLGS, Pth, for each day
from its initial start-up until it was shut down for refurbishment was provided by measurements taken by PLGS
staff. The uncertainties on these power measurements
were within 1%. The recoverable energy per fission, Erec,
was estimated to be 205.9 ± 0.6 MeV, assuming a mid-life
burn-up percent fuel composition of 0.4% U-235, 99.4%
U-238, 0.2% Pu-239 and 0.006% Pu-241. The average number of neutrons emitted per fission, ν , was estimated to
be 2.83 ± 0.01 [11]. The values of  σ for each steel reactor component were obtained from knowledge of the
volumes of the steel parts, the neutron flux tallies associated with each steel part, and estimates of the cobalt impurity concentrations in the steel. As will be demonstrated
in the next section, the uncertainties in the initial cobalt
concentrations were the dominant sources of uncertainty
in the simulated results. Armed with the estimated γ-ray
emission rates of the steel reactor components, we then
used them to convert the γ-ray dose rate tallies into total dose rates for comparison with the measured data.
7. Results of the MCNP5 Simulations
32
Figure 4 contains plots of the simulation results obtained
with the analysis method described in the Section VI. The
measured dose rate data are also included in the plots for
comparison. The initial estimated 59Co concentrations that
were used to obtain the results in Figure 4 and Figure 5
were: 1) 1400 ppm (adjuster rods), 2) 560 ppm (mid-calandria shell), 3) 84 ppm (calandria sub-shells), 4) 84 ppm
(calandria rings), 5) 119 ppm (calandria tubesheets) and 6)
700 ppm (nozzles). Since the cobalt concentrations are not
well known, these cobalt concentrations were obtained by
adjusting them until the simulated dose rate data matched
Figure 4
Top: Overlaid plots of the measured and simulated dose
rates for the adjuster rods at their estimated and designed
retracted positions with error bars on the simulated data.
Bottom: Overlaid plots of the measured and simulated dose
rates for the adjuster rods at their estimated and designed
retracted positions without error bars on the simulated data.
aecl Nuclear Review
vol 1, Number 1, june 2012
measurements of the high dose rate profiles inside a shutdown candu reactor
c. jewett, j. chow, d. comeau, g. jonkmans, b. smith, b. sur, d. taylor and s. yue
The primary objective of performing the MCNP5 simulations was to determine the positions of Adjuster Rods 15
and 16 more accurately. This entailed aligning the simulated dose rate peak due to the calandria and nozzles with the
measured dose rate peak, so that the simulated and measured curves would match each other as closely as possible.
When compared with the MCNP5 geometry, it was found
that the location of the 1079 Gy/hr peak in the dose rate
data is 3421 mm above the centre axis of the main calandria
shell. This is 97 mm below the inner calandria edge, which
is 3518 mm above the centre of the calandria. This is due
to the nearly 80 mm radius VP2 hole in the calandria shell.
This hole results in a diminished dose rate at the calandria
edge, where VP2 meets the calandria shell. The position error with which this alignment was made is estimated to be
± 30 mm, since that is the distance by which one can shift
the alignment in the plots and still obtain a reasonable fit.
Fig. 5 contains re-scaled plots of the dose rates obtained by
simulating the adjuster rods 480 mm and 400 mm below
their designed positions (MCNP5 Estimated Pos. 1 and Pos.
2). As one can see from the graph in this figure, the “Pos.
1” data and “Pos. 2” data bracket the measured dose rates
associated with the adjuster rods. Thus, the estimated uncertainty with which the positions of the adjuster rods were
simulated is half the distance between Pos. 1 and Pos. 2 or
± 40 mm. By adding the squares of the errors on the position measurement (27 mm), the alignment (30 mm), and
simulated position (40 mm), we obtained a total error on
the positions of the adjuster rods of 57 mm. The average
of Pos. 1 and Pos. 2 is 440 mm, and hence, the conclusion is
that the adjuster rods were 440 ± 60 mm below their fully
retracted designed positions at the time of the dose rate
profile measurements. Fig. 5, which includes the dose rate
data obtained by simulating the adjuster rods at their designed positions, provides confirmation of the fact that the
adjuster rods were out of position.
8. Conclusions
The γ-ray dose rate profile of a CANDU reactor that was
undergoing refurbishment was measured for the first time
in 2009. This measurement employed a small, inexpensive silicon diode sensor and a scan drive system. These
measurements produced a γ-ray dose rate profile that revealed the locations of the steel features of the PLGS reactor.
The maximum reactor γ-ray dose rate was measured to be
1079±36 Gy/h. Based upon MCNP5 simulations, this dose
rate maximum occurred about 90 mm below the calandria
edge.
®
Finally, this work demonstrated the fact that by combining
the detector and scan drive system measurement technique
with MCNP5 simulations, it is possible to obtain a onedimensional image of the interior of a shut down reactor,
based upon its γ-ray dose rate profile. Future experiments
are being proposed to evaluate the fully inserted, in-core
design positions of the adjuster rods by performing additional γ-ray profile scans.
Figure 5
Top: Overlaid plots of the measured and simulated dose
rates due to Adjuster Rods 15 and 16. The Position 1
data (MCNP5 Estimated Pos. 1) were created by simulating the adjuster rods at their estimated locations, 480
mm below their designed positions. The Position 2 data
were generated by simulating the adjuster rods 400 mm
below their designed positions. Bottom: Overlaid plots
of the simulated dose rate data and the measured dose
rate data in the region associated with the adjuster rods.
MCNP5 simulations helped to confirm the fact that the adjuster rods were out of position with respect to their fully
retracted, out-of-core state. By combining the measured
and simulated data, it was determined that the adjuster rods
were off from their out of core position by 440 ± 57 mm.
AECL NUCLEAR REVIEW
33
aecl Nuclear Review
vol 1, Number 1, june 2012
References
measurements of the high dose rate profiles inside a shutdown candu reactor
c. jewett, j. chow, d. comeau, g. jonkmans, b. smith, b. sur, d. taylor and s. yue
[1] X-5 Monte Carlo Team, 2003, MCNP — A General Monte Carlo N-Particle Transport Code, Version 5, Volume I: Overview and Theory, LA-UR-03-1987, April 24, (Revised 2/1/2008)
[2] B. Sur, S. Yue, and A. Thekkevarriam, 2007, “Radiation Exposure Rate and Liquid Level Measurement Inside a High Level Liquid Waste (HLLW) Storage Tank”, 28th Annual CNS Conference, Saint
John, New Brunswick, Canada
[3] B. Sur, S. Yue and G. Jonkmans, 2009, “A Detector System for Measuring High Radiation Fields”, Sixth American Nuclear Society International Topical Meeting on Nuclear Plant Instrumentation,
Control, and Human-Machine Interface Technologies, Knoxville, Tennessee, USA
[4] S. Yue, et al., 2010, “Measurement of a Gamma Radiation Profile inside a CANDU® Reactor”, Proceedings of the 18th International Conference on Nuclear Engineering, Xi’an, China
[5] G.F. Knoll, 1989, Radiation Detection and Measurement, 2nd Edition, John Wiley & Sons
[6] Keithley Instruments, 2002, Model 6487 Picoammeter/Voltage Source Reference Manual, Doc. No. 6487-901-01 Rev. A, Keithley Instruments, Inc. Cleveland, OH, USA
[7] P.M. Mathew, et al., 2009, “Severe Core Damage Accident Analysis for a CANDU Plant”, OECD/NEA Workshop, “Implementation of Severe Accident Management Measures”, Schloss Bottstein,
Switzerland
[8] Allegheny Technologies Wah Chang, 2003, “Technical Data Sheet: Reactor Grade Zirconium Alloys for Nuclear Waste Disposal”
[9] Weisstein, Eric W. “Maximum Likelihood.” From MathWorld--A Wolfram Web Resource. http://mathworld.wolfram.com/MaximumLikelihood.html
[10] John R. Lamarsh, 1966, Introduction to Nuclear Reactor Theory, Addison-Wesley Publishing Company, Inc.
[11] V. I. Kopeikin, et al., 2004, “Reactor as a Source of Antineutrinos: Thermal Fission Energy”, Physics of Atomic Nuclei, Vol. 67, No. 10, pp. 1892-1899
34
FULL ARTICLE
Abstract
For programs that solve the neutron
transport equation with an approximation
that the neutron flux is constant in each
space in a user-defined mesh, optimization
of that mesh yields benefits in computing
optimization of the spatial mesh
for numerical solution of the
neutron transport equation in a
cluster-type lattice cell
R.S. Davis*
Atomic Energy of Canada Limited, Chalk River Laboratories, Chalk River, Ontario, Canada, K0J 1J0
test runs of the solving software would
Article Info
Keywords: neutron transport equation, mesh optimization, cluster lattice cell
Article history: Received 15 April 2012, Accepted 25 June 2012, Available online 30 June 2012.
*Corresponding Author: (613) 584-3311 ext. 44278, [email protected]
be necessary. The method presented here
1. Introduction
time and attainable precision. The previous
best practice does not optimize the mesh
thoroughly, because a large number of
optimizes the mesh for a flux that is based
on conventional approximations but is more
informative, so that a minimal number
of parameters, one per type of material,
must be adjusted by test runs to achieve
thorough optimization. For a 37 element,
natural-uranium, CANDU® lattice cell,
the present optimization yields 7 to 12
times (depending on the criterion) better
precision than the previous best practice in
37% less computing time.
Many computer programs that solve the neutron transport equation require the
input data to specify a computational mesh that partitions the space to be analyzed,
and make the approximation that in each thus-defined mesh space the neutron flux
(hereafter just called “flux”) is constant as a function of position. The choice of the
mesh structure directly affects the speed and precision of the resulting transport
calculations. “Precision” in this context means agreement between the solution
that the program calculates and an exact solution of the continuous transport equation that the program represents in its discretized approximation. Such precision
is one of the requirements for the accuracy of a mathematical model; examples of
other requirements are accuracy of physical data and conformance of the continuous equation to the actual physics. This statement supposes that imprecision is
not to be deliberately introduced to cancel out inaccuracy from other sources, i.e.
no phenomenological fitting. This paper describes a method of optimizing such a
mesh, so that it yields more precise results using fewer mesh spaces than with earlier optimization methods, with particular application to a cluster-type lattice cell.
One purpose that mesh optimization can serve is economy; the mesh is to yield
the best compromise between computing time, which depends primarily on the
number of mesh spaces, and the precision of the results, which depends on their
fineness. This goal is important if the mesh is to be used many times; for example,
if a mathematical model that uses the mesh is to be validated for design and operation of nuclear facilities.
Another purpose that mesh optimization can serve is maximum precision. Finer
mesh spaces reduce discretization error, but roundoff error increases with the
number of mesh spaces, so that mesh optimization is also necessary for maximum
precision. Since an exact solution of the transport equation is not available, a reference solution with maximum precision is necessary to gauge the precision of each
mesh that is proposed for optimization for economy.
Optimization of a mesh involves test runs with varied meshes to learn the resulting precision and the execution time for each proposed mesh. An infinite variety
of meshes can represent any one physical configuration, but practicality limits the
number of test runs. Consequently, optimization requires inference techniques
that can identify a practicably small subset of all possible meshes, with assurance
that the optimum mesh will be in that subset. This paper describes an improved
technique, which identifies a smaller such subset of the meshes that describe a
cluster-type lattice cell than previous common practice, described below, identifies. As illustration, this paper uses it to optimize a two-dimensional representation of a 37 element, natural-uranium, CANDU® lattice cell [1, 2, 3, 4, 5] in the lattice
AECL NUCLEAR REVIEW
35
aecl Nuclear Review
vol 1, Number 1, june 2012
optimization of the spatial mesh for numerical solution of the neutron
transport equation in a cluster-type lattice cell - r.s. davis
code WIMS AECL [6], and compares that mesh with a mesh
that was optimized for the same purpose, configuration,
and software by previous common practice. The improved
technique yields a greatly different mesh, which yields considerable gain in precision and somewhat reduced computing time compared with previous best practice.
2. Basic Previous Methods
The algorithm derived here that performs part of the optimization of a mesh is based on a technique of adaptive
meshing that underlies most related previous work [6, 7,
8] as is shown below: since the flux in each mesh space is
approximated to be constant, the imprecision incurred by
the use of a proposed mesh is an increasing function of the
greatest range, i.e., maximum minus minimum, of the actual flux in a mesh space. Thus, part of the optimization
becomes the well-defined condition of having equal range
of flux in each mesh space. Use of the actual flux in this
measure would be impractical, but optimization can be
achieved with an approximation of the actual flux.
Although this basic concept is widely used, there seems to
be no standard name for the approximate flux thus used; it
is hereafter called the “mesh-criterion flux”. It is denoted

f (r ) ; the origin of coordinates is the center of the fuel chan
nel. The equation
 for f (r ) differs, depending on what part
of the cell is at r . Its units cancel out in use, so it is here
considered unitless.
The optimization used by the cited and other work follows
from the approximations that inside a cylindrical region of
approximately uniform neutronic properties

2
f (r ) = a1 ρ1 (r ) + b1
(1)
and outside a cylinder, in a surrounding region with properties contrasting the cylinder,


f (r ) = a2 log(ρ 2 (r )) + b2

(2)
in which ρ(r ) is distance from the central axis of the cylinder. These approximations follow from the diffusion equation in the limit of small distances [9].
The symmetry of these expressions implies that each such
region is to be subdivided into annular mesh spaces with
which to represent the discretized flux. Thus, the abovestated condition of equal range of flux in each mesh space
is that


 max ( f (r )) −  min ( f (r ))


r
Region
∈
r ∈Region
(3)
 max ( f (r )) −  min ( f (r )) =
r ∈Annulus
r ∈Annulus
N Annulus
36
in which NAnnulus is the number of annuli used in a cylindrically bounded region.
The approximations in Equations (1) and (2) are realistic
only if the coefficients a and b depend on neutron energy,
but mesh-based transport-analysis software uses the same
mesh at every energy. Equation (3) resolves this paradox; it
yields a set of annuli that is independent of the coefficients
a and b (with a removable singularity at a = 0 ), and thus
has the necessary independence of energy. More sophisticated approximations would lack this property. With Equations (1) and (2), Equation (3) yields the results, found in
the cited and much other work, that inside a cylinder the
annuli have equal area, and outside an isolated cylinder the
annuli have equal ratios of inner to outer radius.
The weakness of the approximations in Equations (1) and
(2) is that they are appropriate only for isolated cylinders.
Multiplicity of cylinders, such as fuel rods in a bundle and
fuel channels in a lattice, causes azimuthal variation of flux
and different fluxes in different fuel rods, for which those
equations give no estimate. Proper account of these additional variables without a more informative mesh-criterion
flux would call for an impractical number of test runs. Most
related previous work, here called “previous common practice”, uses an equal number of annuli in every fuel rod, an
equal number of sectors in every annulus (except where
sectors are obviously redundant because of symmetry), and
an equal angle spanned by every sector in an annulus [6, 7,
8]. Even with these simplifications, additional test runs are
required to decide the numbers of sectors, and yet the thusderived mesh is not optimal, as is shown in Section 5.
3. Present Method
3.1 Basic Idea
The method of optimization presented here uses a meshcriterion flux that is based on the conventional approximations in Equations (1) and (2), but that consists of more
informative linear combinations of those expressions. It
approximates the azimuthal dependence of the actual flux
and relates fluxes in different parts of the lattice cell to the
greatest extent possible with such approximations. An algorithm then derives an optimum mesh from that meshcriterion flux, guided by parameters that must be adjusted
by test runs. This approach reduces the number of parameters that must be adjusted by test runs to one parameter
for each distinctive category of material, i.e. coolant, fuel,
fuel sheaths, each material in the fuel channel (typically
pressure tube, annular void, and calandria tube), and moderator. This seems to be the minimum set of parameters
that must be determined by test runs, because the actual
flux depends on the characteristics of these materials in a
manner that requires transport calculations.
The parameters to be thus adjusted may be defined in
various ways; those used here are interpretable as numbers
of mesh annuli, hereafter denoted Ncategory, that are to be
aecl Nuclear Review
vol 1, Number 1, june 2012
optimization of the spatial mesh for numerical solution of the neutron
transport equation in a cluster-type lattice cell - r.s. davis
used in each of the above-listed categories of material, with
dimensions that occur in the actual problem. This sort of
definition facilitates human judgement in the selection
of test runs, because it is familiar to users of the sort of
transport software here considered. As is shown below, all
the detailed parameters that characterize a mesh, denoted
Ncategory with additional subscripts, are uniquely determined
by a set of values of Ncategory for the above-listed categories,
so that only the set of values of Ncategory has to be selected by
test runs.
Because fuel composition changes according to the flux in
it, fuel is represented as a distinct material in each mesh
space (except for amalgamation among mesh spaces that
have equal fluxes because of symmetry), with uniform initial composition. The parameters Ncategory thus determine
not only the representation of the flux, but also the representation of the fuel materials.
This basic idea is applied below to a cluster-type lattice cell,
such as in a CANDU lattice.
3.2 Structural Materials
Structural materials, i.e., fuel sheaths and fuel-channel components, are designed to have minimal effect on the neutron
flux by choice of materials and by minimization of quantity. Consequently, representation of azimuthal variation in
them has proven unnecessary if the fuel bundle is centered
in the fuel channel [6, 7, 8]. Less symmetric configurations
have not yet been analyzed by the methods described here.
Because flux in structural materials is dominated by regions
they surround, radial variation is represented in accordance
with Equations (2) and (3). This procedure is no different
from the previous common practice; one parameter for the
mesh in each structural material is to be adjusted by test
runs.
3.3 Lumen of Fuel Channel
3.3.1 Materials
To define the mesh-criterion flux in the lumen of the fuel
channel, the materials to consider are coolant and fuel. Consequently, two parameters, Ncoolant and Nfuel , are used in the
definition of the mesh-criterion flux in the lumen, and are

to be chosen per test runs. In the following, f (r ) is defined
using terms with a range of 1 over a cylinder containing one
category of material, and each such term is multiplied by
either Ncoolant and Nfuel. A maximum value of 1 is imposed on
the range of the mesh-criterion flux in every mesh space in
the lumen. This combination of normalization and condition has the effect that Ncoolant and Nfuel can be interpreted as
numbers of annuli. A direct illustration of this feature follows Equation (5).
A consequence of this normalization is that in any subset of
the lumen, the required number of mesh spaces is the Ceiling of the range of the mesh-criterion flux in that subset, in
which Ceiling is a function that returns the smallest integer
greater than or equal to the argument.
Since the fuel is in cylinders surrounded by the contrasting material coolant, and the lumen, containing fuel and
coolant, is cylindrical and is surrounded by the contrasting
materials fuel channel and moderator, the mesh-criterion
flux in the lumen is based on adaptations of Equation (1) to
these cylinders.
3.4 Mesh Optimization in Coolant
The mesh-criterion flux used in the coolant is, by adaptation of Equation (1),


f (r ) = f Coolant (r ) for
in which
(r ∈ lumen ) ∧ (r ∉ fuel rod ),
r2

f Coolant (r ) = N Coolant
2
RCoolant Outer
− RCoolant Inner
− RCoolant Inner
2
2
+ f (RCoolant Inner ),
(4)
(5)
RCoolant Outer is the inside radius of the fuel channel, and
RCoolant Inner is the radius of the central rod of the fuel

bundle if there is one,
0 if not. r is the magnitude of r .


and is equal to ρ(r ) . f (rR)Coolant Inner is defined by Equation
(6) below as the mesh-criterion flux at the surface of
the central fuel rod if any, 0 if not. Because the range of


f (r ) in the coolant is NCoolant and the range of f (r ) in each
mesh space is limited to 1, NCoolant is the number of annuli.
Azimuthal dependence in coolant does not need to be represented, because the fuel rods are packed closely in comparison with mean free paths in the coolant, and symmetrically [6, 7, 8]. The resulting mesh in the coolant is identical
to the previous common-practice mesh, per Equation (3),
but Equation (5) has further importance because of its use
in the following section.
Because of the small neutronic effect of fuel sheaths, the
mesh-criterion flux is defined to be continuous at the
boundaries that would exist between fuel and coolant in the
absence of the sheaths. Thus, even though Equations (4),
(6), and (7) have domains in different materials, they are
interrelated by this condition of continuity at the boundaries between their domains. Equation (5) includes constant
terms whose purpose is to satisfy this continuity condition, even though they have no effect in the derivation of
the mesh by Equation (3). In related equations below, this
condition of continuity is practically important; adherence
to it in Equation (5) maintains logical consistency.
AECL NUCLEAR REVIEW
37
optimization of the spatial mesh for numerical solution of the neutron
transport equation in a cluster-type lattice cell - r.s. davis
aecl Nuclear Review
vol 1, Number 1, june 2012
3.5 Mesh Optimization in Fuel
3.5.1 Mesh-Criterion Flux in Fuel
The mesh-criterion flux in the central rod of the fuel bundle,
if there is one, or that otherwise would be considered to apply in the innermost rod of the fuel bundle if it were central,
is, by adaptation of Equation (1),

f (r ) = N Fuel
r2
RInner Fuel
2
for
(r ∈ central fuel)
(6)
in which RInner Fuel is the radius of the fuel in the fuel rod closest to the cell’s center, which is the central fuel rod if one
exists. The numbers of annuli used in different fuel rods depend on NFuel through Equations (7) and (8), which contain
expressions that do not generally yield integer values even
when NFuel is an integer. For this reason, a non-integer value
of NFuel will in general result from the optimization, and the
number of annuli used in a central rod is Ceiling(NFuel). If
there is no central fuel rod, then Equation (6) only serves as
a definition of the parameter NFuel by stating how many annuli the innermost rod would have if it were central.
In a non-central cylinder of fuel, the mesh-criterion flux is
the sum of Equations (5) and (6), applied at the location of
the cylinder,
(
)
2
 
2
r − PFuel j − RFuel j


f (r ) = f Coolant (r ) + N Fuel
2
RInner Fuel

for
(r ∈ fuel cylinder number j ) (7)
in which PFuel j is the center of the fuel rod and RFuelj is its
radius. This sum satisfies the condition of continuity of
flux between coolant and fuel, and it thus represents the increase in the gradient of the flux in the fuel that is caused by
proximity to the moderator [6]. Because RFuelj is constant,
its inclusion in Equation (7) has no effect on the derived
mesh; but it yields continuity between fuel and coolant,
which is essential to the logic. Because the two terms in
Equation (7) are multiplied by the two independent quantities NCoolant and NFuel , the sum does not represent the actual
flux in any energy group. Rather, a high value of NFuel leads
to a correspondingly exaggerated gradient of the flux in the
fuel. This yields correspondingly enhanced fineness of the
mesh in the fuel.
The use of RInner Fuel2 in Equation (7) implies that the parameter a in Equation (1) is the same in fuel rods of different
radii and that the number of annuli necessary in a central
rod depends quadratically on its radius, which is a reasonable approximation if their initial compositions are the
same. The present process has not been applied with fuel
rods of different initial compositions; with strong differences, distinct values of NFuel for each initial composition would
presumably be necessary for good optimization.
38
Thus, geometry and two quantities chosen per test runs,
NCoolant and NFuel, completely determine the mesh-criterion
flux in every fuel rod and in the coolant. That and the condition of a maximum range of 1 in each mesh space then completely determine the mesh in fuel and coolant as follows.
3.6 Optimization of Annuli in Fuel
The restriction to a range of 1 in each annulus and the dif
ference in f (r ) between the center of a cylinder of fuel and
its surface nearest the moderator yields the number of annuli in fuel cylinder number j,
((
) (


N Fuel j = Ceiling f PFuel j + RFuel j PˆFuel j − f PFuel j
))
(8)
in which PˆFuel j is a unit vector
 pointing away from the center of the fuel channel, i.e. PFuel j PFuel j , and the condition of
uniform range of mesh-criterion flux implies that the outside radius of the kth annulus in fuel cylinder number j is
(
) (
) ((
) (




ρ Fuel j ,k : f PFuel j + ρ Fuel j ,k PˆFuel j − f PFuel j = f PFuel j + RFuel j PˆFuel j − f PFuel j
in which 1 ≤ k ≤ N Fuel j .
)) N k
Fuel j
(9)
Figure 1C shows in a particular example (described below)
how the annuli become finer as fuel rods are farther offcenter, in addition to the conventional trend toward thinner
annuli toward the outside in each fuel rod.
(a)
(b)
(c)
(d)
(d)
Figure 1
Derivation of mesh from mesh-criterion flux. (a) annulus in
fuel, divided along contours of mesh-criterion flux in coolant; (b) approximation of division in above figure with sectors; (c) optimum mesh in the lumen for 37-Element example; (d) surface plot of mesh-criterion flux in moderator in
one quadrant of a cell; (e) contour plot and optimum mesh
for above example, in moderator.
aecl Nuclear Review
vol 1, Number 1, june 2012
optimization of the spatial mesh for numerical solution of the neutron
transport equation in a cluster-type lattice cell - r.s. davis
3.7 Optimization of Sectors in Annuli Fuel
The second term in Equation (7) is invariant under rotation
about the center of the cylinder; all the azimuthal dependence in Equation (7) is in the first term, which is identical
to Equation (5). In the optimization of sectors within the

above-described annuli, f (r ) may be defined by either of
those equations with the same result. For simplicity, Equation (5) is used here.
The required number of angular subdivisions in the kth annulus is determined by the range of the azimuthal variation
of the mesh-criterion flux in that annulus. Thus,
((
) (


N Fuel j ,k = Ceiling f PFuel j + ρ Fuel j ,k PˆFuel j − f PFuel j − ρ Fuel j ,k PˆFuel j
))
(10)
If N Fuel j ,k ≤ 1 , the annulus is not divided into sectors. Otherwise, Figure 1A shows an exemplary annulus with
N Fuel j ,k = 5 , subdivided along contours of Equation (5), which
occur at uniform steps of the value of Equation (5). However, WIMS AECL and similar software subdivides annuli
along azimuthal boundaries, into sectors. Consequently,
the subdivision in Figure 1a is approximated by the sectors
exemplified in Figure 1b, such that each defining contour of
Equation (5) intersects a sector boundary at the root-meansquare radius of the annulus,
(
2
ρFuel j ,k = ρ Fuel j ,k + ρ Fuel j ,k −1
Thus, the angles are
(
((
)
)
2 12
)) (
)


θ Fuel j ,k ,l : f PFuel j + ρFuel j ,k R θ Fuel j ,k ,l PˆFuel j − f PFuel j − ρ Fuel j ,k PˆFuel j =
( f (P
Fuel j
) (

+ ρ Fuel j ,k PˆFuel j − f PFuel j − ρ Fuel j ,k PˆFuel j
))
l
N Fuel j ,k
(11)
in which R (θ ) is a rotation operator through the azimuthal angle θ , and 0 < l < N Fuel j ,k . For each such l, two symmetrically related solutions θ exist, so there are 2 N Fuel j ,k − 2
sectors. Figure 1C shows in a particular example how the
mesh boundaries between sectors conform in their spacing,
but do not necessarily coincide, with mesh boundaries in
the coolant.
A pair of sectors that are defined by the same values of l will
have equal fluxes by symmetry, and therefore may be defined to have the same material, as is illustrated by the colors
in Figures 1a and 1b. Although an annulus has 2 N Fuel j ,k − 2
sectors, only N Fuel j ,k different fluxes occur in them. In the
example in Figure 1c, the fuel is represented as 50 different
materials, though with uniform initial composition.
3.8 Moderator
3.8.1 Mesh-Criterion Flux in Moderator
The mesh-criterion flux in the moderator is defined to satisfy reflective or periodic boundary conditions on the cell
boundaries, by supposing that it is in an infinite lattice. Departures from such boundary conditions are managed independently of the optimization process, as is described in
Section 5. The mesh-criterion flux is

f (r ) = − 
∑(
PChannel j ∈K N )
(log(r − P
2
Channel j
))+ s tan2(ππ rs ) P
1
2
for
(r ∈ moderator ) (12)
The origin of coordinates is the center of the cell. P1 is the
distance from there to the center of a nearest neighbor.
K(N) is the set of locations of centers of fuel channels at distances ≤ N P1 from the center of the cell of interest (including distance 0); the x axis points to a nearest neighbor. s is
the number of sides of each cell, e.g., four for a square lattice. The first term represents the mesh-criterion flux in the
moderator due to each cell with a channel in K(N) per Equation (2). The second term represents the infinite number of
other cells approximated as a continuous medium, which
yields a flux in the cylinder occupied by K(N) per Equation
(1). Equation (12) has no particular normalization; there
is no simplification analogous to the normalization of the
mesh-criterion flux in the lumen because of the weighting
in Equation (13).
Equation (12) only approximately satisfies the boundary
condition of zero gradient perpendicular to the cell’s bound
aries. The coefficient of r 2 is determined by the condition that the perpendicular gradient integrated around the
cell’s boundaries is 0. The approximation improves, i.e., the
maximum magnitude of perpendicular gradient across the
boundary decreases, as N increases. The set K(3) is recommended; it has only 29 members in a square lattice, but it
satisfies the reflective boundary condition with precision,
defined by (maximum magnitude of gradient across bound
ary linearly extrapolated across cell) / (range of f (r ) in
moderator), better than 0.1%. Equation (12) with K(3) for
one quarter of a square lattice cell is shown as a surface plot
in Figure 1d, and as a contour plot in Figure 1e.
3.8.2 Optimization of the Mesh in the Moderator
The number of annuli used in the moderator is NModerator,
which is another parameter to be optimized by test runs.
As a function of the magnitude,
 r , the mesh-criterion flux
has its greatest range when r . is directed along a diagonal,
e.g. in the direction π s . The mesh in the moderator cannot be optimized by setting the radii of the annuli and then
assigning sectors in each, as elsewhere in the cell, because
of the more complicated dependence of the mesh-criterion
flux on position. However, in the procedure used here, an
initial assignment is made, that annulus j has outside radius


rModerator j : f (r (RChannel , π s )) − f r rModerator j , π s = j
(
((
 
 f (r (RChannel , π s )) −


)))
 

P1
f  r 
, π s   
  
  2 cos(π s )
∑w
k
k =1
N Moderator
∑w
k =1
k
AECL NUCLEAR REVIEW
(13)
39
optimization of the spatial mesh for numerical solution of the neutron
transport equation in a cluster-type lattice cell - r.s. davis
aecl Nuclear Review
vol 1, Number 1, june 2012

in which the positions r . are expressed in polar coordinates of radius and angle. RChannel is the inside radius of the
moderator. The weights, wk , are all 1 for this initial assignment.
Sectors are assigned to each annulus in accordance with its
radii. Minimization of the number of mesh spaces subject
to a restriction on the range of the mesh-criterion flux is
equivalent to maximization of the area of each mesh space
subject to the same restriction. In a first-order approximation, this maximum occurs when the radial variation and
the azimuthal variation of the mesh-criterion flux are equal.
Consequently, in annulus number j, in each of the 2s symmetrically related sectors, the number of sectors is
((
((
)) (r(r
)) (r(r

 f r rModerator j , θ min j − f
N Moderator j = NInt 
 f r rModerator , π s − f
j −1

))
))
,π s 


Moderator j , π s 
Moderator j
(14)
in which NInt (x) is a function that returns the
integer near2
2 12
rModerator j = rModerator j + rModerator j −1
est x,
(
)
is the root-mean-square radius of moderator annulus j, and
θ min j = cos −1 12 P1 rModerator j
(
)
if the circle of radius rModerator j intersects the cell boundary,
0 otherwise.
If NModeratorj >1 then NModeratorj sectors are defined in each symmetrically related sector. Their bounding angles are such
that the range of the mesh-criterion flux is the same in every sector in annulus number j. Any adjacent two sectors
whose common boundary is one of the 2s lines of reflection symmetry are then merged, because the range of the
mesh-criterion flux is no greater in that union than in its
components.
In accordance with Section 2, the incurred error attributed
to each annulus is the range of the mesh-criterion flux in
each of its sectors if it has them, or in the whole annulus
if not. The criterion of optimization requires all of the attributed incurred errors to be equal. Although the abovedescribed steps are designed for that goal, the attributed incurred errors typically vary, because Equation (14) changes
discontinuously as a function of the properties of an annulus. Worst is the transition in consecutive annuli from
no sectors to two sectors, which changes the attributed incurred error by almost a factor 2.
40
As a partial remedy, the radii of the annuli are then reassigned per Equation (13), but this time weight is the inverse
of the incurred error attributed to annulus number k. Then,
the above-described assignment of sectors per Equation
(14) is repeated. This reduces the range of variation of the
attributed incurred errors typically to ±20% of the average.
Better apportionment of these weights would yield better
optimization, but further iteration by the above-described
procedure makes the mesh worse, not better, because of the
discontinuous behavior of Equation (14).
4. Tracking Lines
In software that uses collision probabilities, such as WIMSAECL, most of the computing time is consumed in preparing
a matrix of collision probabilities. This time depends on the
number of tracking lines, which are used for ray tracing [6],
as well as the fineness of the mesh. For the sake of precision
consistent with the fineness of the mesh, different spacings
are used for different ranges of distance of tracking lines
from the cell center. The spacing of tracking lines in each
radial range is proportional to the thickness of the thinnest
annulus through which each tracking line passes at its closest approach to the center of the lattice cell. The ratio is
denoted NLine, defined such that a larger value means more
lines, and is another parameter to be optimized by test runs.
The number of angles of tracking lines, NAngle, is another
parameter to be optimized by test runs. One value is used
throughout each test run.
5. Results for 37-Element Example
5.1 Choice of Parameters
The above logic yields an optimized mesh, given a set of values of the parameters. The other component of optimization is the choice of that set of values. Even with the more informative mesh-criterion flux, this optimization occurs in a
rather high-dimensioned space. The mesh in structural materials is fairly simple to optimize, and the present method
offers no improvement on previous common practice. Test
runs with the present mesh optimization have confirmed
earlier findings that, for economy, one annulus in each such
material is optimum, and for the highest precision one annulus in each fuel sheath and three annuli in each other
structural material suffice. However, there remain the five
parameters NCoolant, NFuel, NModerator, NLine, and NAngle.
Although sophisticated methods exist for optimization in
multidimensional space with a minimum number of test
runs, for the present work many sets of these parameters
have been given test runs, with the goal of general understanding of the effects of varied choices as well as finding
an optimum.
5.5 Optimization Metrics
Although the mesh-criterion flux derived above closely
resembles plots of actual flux [6], its ultimate test is the
agreement between calculations made with meshes optimized in accordance with it and a reference calculation optimized for precision. The reference calculation
aecl Nuclear Review
vol 1, Number 1, june 2012
optimization of the spatial mesh for numerical solution of the neutron
transport equation in a cluster-type lattice cell - r.s. davis
for the 37 element, natural-uranium, CANDU lattice cell [1,
2, 3, 4, 5] example used here is derived from the parameters
NCoolant = 40, NFuel = 8.2, NModerator = 60, NLine = 4.4, and NAngle =
11.1 annulus in each fuel sheath, and 3 annuli in each fuelchannel component. (This fineness makes a diagram impracticable.)
Two metrics of agreement are used throughout the present work: agreement of k∞ , and agreement of change of
k∞ upon removal of coolant. The latter is hereafter called
“CVR” for “coolant void reactivity”. From each test run, these
quantities are compared at a predetermined set of five irradiation times that correspond approximately to fresh fuel,
equilibrium short-term poison, maximum k∞ , half-way
thereafter to the following point, and k∞ = 1 Root-meansquares over those five irradiation points of the two metrics
are shown in Figure 2. A few test runs have also gauged
the degrees of agreement in change of k∞ as a function of
fuel temperature and of moderator boron concentration.
Agreements by these criteria are similar to the agreements
in CVR. These optimization metrics are the important ones
for applications thus far made of the present process; it has
not been tested against other optimization metrics, such as
detailed power and flux distributions.
0.30
1.50
1.25
R.M.S. Difference from
Precise Calculation
versus
Computing Time
0.25
Finer and coarser reference meshes have also been tried;
results have nearly the same relative positions as shown
in Figure 2, but with the whole set of discrepancies shifted higher by 0.1 to 0.2 milli k. Thus, the present reference
mesh seems to be near a shallow optimum of precision.
Computing times shown in Figure 2 are expressed as multiples of the computing time on the same computer with a
previous “standard” mesh, which is further described below.
5.3 Optimization
The best possible precisions for different amounts of computing time are the lower envelopes of the two sets of results. Figure 2 shows that greater computing time can be
rewarded with greater precision up to 0.63× the computing time for the previous best-practice mesh, and computing time above that is wasted; a discontinuity in the trend
occurs at that point in the agreements for k∞ , and the
agreements for CVR, although not so clearly patterned, are
consistent with it. The mesh at that transition is labelled
“Optimum” in Figure 2, because it will indeed be optimal for
any purpose other than an overwhelming need for speed.
Its parameters are NCoolant = 10, NFuel = 4.2, NModerator = 15, NLine
= 1.5, and NAngle = 7. It is shown in Figures 1c and 1e.
5.4 Comparison with Previous Common-Practice Optimization
5.4.1 Subject of Comparison
Previous
Optimum
CVR
Previous
Optimum
0.20
CVR, mk
k-inf, mk
1.00
k-infinity
0.75
0.15
0.50
0.10
0.25
0.05
0.00
0.25
0.50
0.75
1.00
Computing Time / Previous Computing Time
0.00
1.25
Figure 2
Root-Mean-Square differences from precise calculation in
milli-k versus computing times relative to previous best
practice for 37-Element example.
The above-described “optimum” mesh is here compared
with a mesh that was previously designed as a “standard”
mesh for the same configuration. Like the “optimum” mesh
derived here, that previous mesh is optimized for a good
compromise of speed with economy, and intended to be
run on the same version of WIMS AECL [6], using presentday computing hardware. The one important difference is
that that mesh was designed according to previous common practice as described in Section 2. Because of these
similarities and differences, that mesh is referred to here as
“previous best practice”. It is shown in Figures 3a and 3b.
5.4.2 Precision
Figure 2 also shows precisions obtained with the previous
best-practice mesh; they are near the upper-right, i.e., least
favorable corner. In 0.63× the computing time for the previous best-practice mesh, the “optimum” mesh yields about
12× better precision in k ∞ and 7× better precision in CVR.
These strong differences correspond with strong contrasts
between the two meshes, which are described below.
5.4.3 Mesh in Lumen
AECL NUCLEAR REVIEW
41
aecl Nuclear Review
vol 1, Number 1, june 2012
optimization of the spatial mesh for numerical solution of the neutron
transport equation in a cluster-type lattice cell - r.s. davis
The respective meshes in the lumen are shown in Figures
1c and 3a. In the coolant, the optimum mesh is identical to
the previous best-practice mesh. In the fuel, the previous
best-practice mesh has the same 4 annuli in each cylinder
of fuel, with 2 sectors in each annulus except in the central
rod, making 8 mesh spaces in each fuel cylinder except for
4 in the central rod. The optimum mesh in the fuel is much
finer and more varied; the consecutive rings of fuel have 5,
6, 6, and 7 annuli, and the numbers of sectors range from 1
to 6. In the outer ring, each cylinder of fuel has 22 subdivisions after deducting for symmetry per Section 3, versus 8
in the previous best-practice mesh there. This contrast is
probably the main reason for improved precision.
5.4.4 Mesh in Moderator
The opposite contrast occurs in the moderator; the respective meshes are shown in Figures 1e and 3b. The previous
best-practice mesh there has there 36 annuli, totalling 434
mesh spaces. The optimum mesh has only 15 annuli, totalling only 62 mesh spaces. The contour bands that underlie
the mesh in Figure 1e show how the apportionment of sectors results from optimization according to the mesh-criterion flux. Although the optimum mesh uses only 1/7 as
many mesh spaces, the narrowest sectors in the optimum
mesh have angular span of only 9.7°, while all the sectors in
the previous best-practice mesh have angular spans of 15°.
This means that most of the sectors in the previous bestpractice mesh have wastefully small ranges of the meshcriterion flux, yet some of those sectors have larger ranges
than in the optimum mesh.
Similar contrast occurs in the apportionment of annuli, and
is due to the representation of reflective boundary conditions in the present mesh-criterion flux. Indeed, test runs
show good precision with NModerator = 12, but a value of 15
allows for deviation from reflective boundary conditions
in a multicell case. Similarly, although the above-described
optimization yields a particularly large mesh space in each
corner of a square cell as shown in Figure 1e, that mesh
space is divided into two radially for better representation
in a multicell case.
6. Conclusion
For neutron-transport software that divides the analyzed
configuration into mesh spaces and makes the approximation of constant flux in each such space, optimization of the
mesh can have great effects on the accuracy of the solution
and on the execution time. For a lattice cell with cluster
geometry, inference of an optimum mesh from the meshcriterion fluxes given by Equations (7) and (12), in comparison with the previous best practice based on Equations
and (2), enables an optimum mesh to be formed with fewer
test runs of the solving software, and produces a mesh that
42
requires less computing time but produces a considerably
more precise result for the integral parameters against
which it has been tested. This conclusion has not yet been
tested against detailed parameters, such as power and flux
distributions.
Improved precision stems mainly from informed, finer subdivision in fuel rods, emphasizing those that are farther offcenter, while improved execution time stems mainly from
informed reduction in the number of subdivisions in the
moderator. However, the latter reduction can degrade flux
mapping in the moderator.
The observed improvements stem ultimately from use of a
more informative mesh-criterion flux. In other transport
calculations, optimization by means of similarly informative mesh-criterion fluxes may also improve precision and/
or reduce computing time.
7. Acknowledgements
The author thanks the following colleagues in AECL for
their knowledgeable help in the work described and in the
preparation of this paper: F. Adams, N. Alderson, D. Altiparmakov, J. Pencer, S. Pfeiffer, D. Roubtsov, A. Trottier. The author is solely responsible for any errors of fact or deficiencies in expression that remain in spite of their help.
(a)
(b)
Figure 3
Previous best practice mesh; (a) previous best practice lumen; (b) previous best practice moderator.
aecl Nuclear Review
vol 1, Number 1, june 2012
optimization of the spatial mesh for numerical solution of the neutron
transport equation in a cluster-type lattice cell - r.s. davis
References
[1] International Atomic Energy Agency, Vienna, June 1996, “In Core Fuel Management Benchmarks for PHWRs”, IAEA-TECDOC-887
[2] International Atomic Energy Agency, Vienna, April 2002, “Heavy Water Reactors: Status and Projected Development”, Technical Reports Series No. 407
[3] K.S. Kozier, 2002, “Assessment of CANDU Reactor Physics Effects Using a Simplified Whole-Core MCNP Model”, PHYSOR 2002 Conference, Seoul, Korea
[4] M.A. Lone, 2001, “Fuel Temperature Reactivity Coefficient of a CANDU Lattice Numerical Benchmark of WIMS AECL (2 5d) Against MCNP”, 22nd Annual Conference of the Canadian Nuclear
Society, Toronto, Canada
[5] J.M. Pounders, F. Rahnema, D. Serghuita, and J. Tholammakkil, 2011, “A 3D stylized half-core CANDU benchmark problem”, Annals of Nuclear Energy, 38(4), pp. 876-896
[6] D. Altiparmakov, September 2008, “New Capabilities of the Lattice Code WIMS AECL”, PHYSOR 2008 Conference, Interlaken, Switzerland
[7] G. Harrisson and G. Marleau, April 2012, “Computation of a Canadian SCWR Unit Cell with Deterministic and Monte Carlo Codes”, PHYSOR 2012 conference, Knoxville, Tennessee, USA
[8] W. Shen, September 2006, “Development of a Multicell Methodology to Account for Heterogeneous Core Effects in the Core-Analysis Diffusion Code”, PHYSOR 2006 conference, Vancouver, British Columbia, Canada
[9] M. Abramowitz, I.A. Stegun, 1972, “Handbook of Mathematical Functions: with Formulas, Graphs, and Mathematical Tables”, Dover Publications, New York
AECL NUCLEAR REVIEW
43
44
FULL ARTICLE
Abstract
Self-excited acoustic resonance is a design
concern in many engineering applications
such as tube bundles of heat exchangers
and boilers. Since this phenomenon is not
yet fully understood, it can be dangerously
unpredictable. Due to the complexity of
the flow-sound interaction mechanisms
self-excited acoustic resonance of
isolated cylinders in cross-flow
A. Mohany*
University of Ontario Institute of Technology, Faculty of Engineering and Applied Science, Oshawa, Ontario, Canada, L1H 7K4
Article Info
Keywords: acoustic resonance, vortex shedding, single cylinder, two-tandem cylinders, two side-by-side
cylinders.
Article history: Received 14 April 2012, Accepted 25 June 2012, Available online 30 June 2012.
*Corresponding Author: (613) 584-3311 ext. 5720, [email protected]
in tube bundles, the simplified cases of a
single cylinder and two cylinders in various
arrangements, tandem and staggered, are
investigated in some detail. A summary
of these investigations is presented in
the current paper. It is found that the
aeroacoustic response of two-tandem and
side-by-side cylinders in cross-flow can be
considerably different from that of a single
cylinder under similar flow conditions.
Moreover, for the case of two tandem
cylinders, the acoustic resonance is excited
over two different ranges of flow velocity;
the pre-coincidence and the coincidence
resonance ranges. The pre-coincidence
acoustic resonance phenomenon is found
to be similar to the acoustic resonance
mechanism of in-line tube bundles.
I. Introduction
The vortex-shedding phenomenon has been a subject of research since it was observed by Leonardo da Vinci. Vortex shedding occurs due to the boundary layer
separation around bluff bodies in cross-flow. The boundary layer separates into
two shear layers that trail and roll-up in the near field forming a periodic vortex
street. In the case of a duct containing a bluff body such as a circular cylinder, when
the vortex shedding frequency coincides with one of the acoustic natural frequencies of the duct, a feedback cycle may occur where the vortex shedding acts as a
sound source and excites an acoustic standing wave which, in turn, enhances the
shedding process and thereby creates a strong acoustic resonance. This process is
known as flow-excited acoustic resonance.
While there has been a large amount of research on the flow around circular
cylinder(s) because of its wide application in engineering practices, see for example
Zdravkovich [1, 2], Igarashi [3, 4], Ljungkrona et al. [5], Lin et al. [6], Alam et al. [7]
and Sumner [8, 9], the flow-excited acoustic resonance phenomenon has received
considerably less attention. Better understanding of this excitation mechanism and
the details of the coupling mechanism and the associated energy transfer between
the flow field and the resonant sound field for the simple cases of a single, tandem, and side-by-side cylinders is necessary in order to explain the phenomenon
of flow-excited acoustic resonance in more complex flow situations such as those
existing in tube arrays. It would also help developing effective means to control and
avoid the occurrence of acoustic resonances.
The flow patterns around two cylinders arranged close to one another in tandem,
side-by-side, or staggered configuration, as shown in Figure 1, are very different
from the case of a single cylinder in cross-flow due to the interference mechanism.
This interference could create hydrodynamic forces that may enhance or suppress
the vortex shedding process. Pressure measurements and flow visualization have
often been used in the literature to identify the interference regions between two
cylinders. Zdravkovich [1] reported that for the case of two tandem cylinders, there
is no existence of vortex shedding behind the upstream cylinder for spacing ratios
(L/D) less than 3.8, where L is the center to center distance between the cylinders and D is the cylinder diameter. Moreover, a “critical spacing” was introduced
to specify the center-to-center spacing at which vortices start to form in the gap between the cylinders. The critical spacing (L/D) was found to be about 3.5, and it is
slightly influenced by Reynolds number. For the case of two side-by-side cylinders
with intermediate spacing ratios, T/D < 2.2, the wakes exhibit bi-stable flow. This is
characterized by a wide-open wake behind one cylinder and a closed-narrow wake
behind the other creating a biased jet between the two cylinders that can switch
its biased direction to either of the two cylinders. However, for large spacing ratios,
T/D > 2.2, the two side-by-side cylinders do not exhibit bi-stable flow properties.
AECL NUCLEAR REVIEW
45
aecl Nuclear Review
vol 1, Number 1, june 2012
self-excited acoustic resonance of isolated cylinders in cross-flow
a. mohany
Instead, the wakes appear more symmetric and will become
dominated by vortex shedding at a single frequency.
The vortex-shedding could be altered by the excited acoustic standing wave, and a significant change to the flow pattern around the bluff body could take place. The acoustic
standing wave forces the flow to oscillate around the bluff
body. The effect of acoustic resonance on the vortex shedding process can be investigated by either oscillating the
flow around a fixed body using an external sound field, or
by oscillating the body itself. Both of these approaches provide controlled experiments to investigate the flow-acoustic
coupling process, and may explain the changes in the flow
patterns and the exerted hydrodynamic forces on the cylinders due to the applied sound. But it is rather difficult to
understand the mechanism of energy transfer between the
flow field and the sound field unless the self-excited resonance case is investigated.
Figure 1
Two cylinders arranged in (a) tandem; (b) side-by-side; and
(c) staggered.
46
Blevins studied the effect of sound on vortex shedding from
a single cylinder [10]. His experiments were performed in
a wind tunnel with and without externally applied sound
field at Reynolds numbers up to 8.5×104. The acoustic
standing wave was excited across the test section using two
electromagnetic speakers. It was observed that the sound
field could correlate the vortex shedding in the span-wise
direction, which makes the shedding process more intense.
Blevins measured the coherence between a flush-film and
a movable probe along the cylinder span with and without externally applied sound [10]. He observed that for
the case of externally applied sound, the vortex shedding
is two-dimensional and the turbulence intensity in the approach flow does not have any effect on the coherent structure of the vortex shedding. Moreover, by applying sound
field at the acoustic resonance frequency of the duct “first
cross mode”, the broadness of the vortex-shedding peak
was reduced as the amplitude of the sound wave was increased. The broad vortex shedding peak was replaced by
a sharp peak at the sound frequency when the sound pressure amplitude was increased to 250 Pa. Blevins investigated this entrainment process by placing the cylinder at
different locations in the transverse direction [10]. He concluded that it is the acoustic particle velocity, not the acoustic pressure that causes the entrainment process.
In the case of oscillating bluff body, whether the oscillations
are self-induced or externally imposed, the lock-in condition can be achieved if the oscillating frequency in the transverse direction is close to the vortex shedding frequency.
Lock-in can also be achieved if the oscillating frequency in
the stream-wise direction is close to twice the vortex-shedding frequency [11]. The lock-in condition in the forced
oscillation has some similarities with the case of acoustic
resonance. One of the early investigations of the wake structure behind a vibrating circular cylinder was performed by
Koopmann [12]. The experiments were conducted in a low
turbulence wind tunnel at Reynolds number ranges of 100,
200 and 300. For the case of a stationary cylinder, threedimensional pattern of the vortex shedding was observed.
However, when the cylinder vibrates at a frequency equal
or close to that of the natural vortex shedding process, the
wake structure behind the cylinder becomes two dimensional, i.e., the vortices along the cylinder span are correlated (shed in-phase) and the vortex filaments are shedding
parallel to the cylinder. Moreover, during this two-dimensional shedding, the upstream turbulence did not have any
effect on the wake structure. This observation is similar to
that observed by Blevins [10] for the case of externally applied sound. Koopmann [12] examined also the conditions
for which the vortex shedding frequency was locked-in to
the oscillation frequency of the cylinder. He observed that
when the cylinder oscillates with a frequency equal to that
of the vortex shedding, a minimum oscillation amplitude of
about 10% of the cylinder diameter is required to initiate
the lock-in mechanism. Nevertheless, the upper and lower ranges of oscillation frequencies for the lock-in region,
around ±25% of the vortex shedding frequency, depend on
the oscillation amplitudes and to some extent depend also
on Reynolds number. Moreover, Koopmann [12] observed
that when the cylinder oscillates at the upper frequency
limit, the longitudinal spacing between the vortex filaments
decreases, and when the cylinder oscillates at the lower frequency limit, the longitudinal spacing between the vortex
filaments increases.
Griffin & Ramberg [13] conducted experiments in an openjet wind tunnel at Reynolds numbers of 144 and 190 to investigate the effect of cylinder vibration on the wake structure of vortex shedding. Velocity measurements in the near
field and flow visualization were performed. They found
that by increasing the oscillation amplitude of the cylinder,
the length of vortex formation decreases. Moreover, the lateral vortex spacing is inversely proportional to the oscillation amplitude, whereas the longitudinal vortex spacing is
inversely proportional to the oscillation frequency, which is
similar to that observed by Koopmann [12].
aecl Nuclear Review
vol 1, Number 1, june 2012
self-excited acoustic resonance of isolated cylinders in cross-flow
a. mohany
Bishop and Hassan [14] investigated the effect of forced oscillation on the hydrodynamic forces, including lift and drag
forces, exerted on a single cylinder. Their experiments were
performed in a water tunnel with a Reynolds number in
the range of 5×103 to 10×103. The investigation focused on
measurements of both the amplitude and the phase angle
of the hydrodynamic forces with respect to the cylinder displacement. The phase angle is important because it influences the mechanism of energy transfer between the flow
field and the cylinder oscillation, which will be discussed in
some details later on. Bishop and Hassan [14] found a remarkable sudden change in both the phase and amplitude
of the hydrodynamic forces at the frequency coincidence.
Many years later, Zdravkovich [15] examined the flow visualizations of previous studies and he noted that for a
reduced velocity smaller than that at the frequency coincidence, the vortex formed on one side of the cylinder is shed
just before the cylinder reaches the position of the maximum amplitude on the opposite side. For a reduced velocity larger than that at the frequency coincidence, the vortex
with the same circulation as before is shed when the cylinder reaches the position of the maximum amplitude on
the same side. Therefore, Zdravkovich [15] explained this
phase jump as a lack in the synchronization between the
cylinder oscillation and the “newly shed vortices”.
For the case of self-excited acoustic resonance, Blevins and
Bressler [16] have performed experiments on the acoustic
resonance in heat exchanger tube bundles. The first phase
of their experiments focused on the aeroacoustic response
of a single cylinder in cross-flow under resonance conditions, and the second phase investigated the aeroacoustic
response of different tube array configurations. The main
objective of these experiments was to develop a model that
can be used to predict the resonant sound pressure level for
a single and multiple cylinders in cross flows.
From the previous investigations, it is clear that a crucial
event in the mechanism of acoustic resonance is the ability
of sound to modulate, and essentially “lock-in”, the process
of vortex shedding. While this phenomenon is relatively
well understood for the case of a single cylinder there are
many unresolved issues for the more complex case of multiple cylinders in close proximity. In the present paper, a
summary of a research program that was conducted on the
flow-sound interaction mechanism of multiple bluff bodies
in cross-flow is presented. Both experimental and numerical results are outlined in the paper.
2. Experimental Set-up
The experiments were performed in an open loop wind tunnel. The test section was made of 25.4 mm thick clear acrylic walls, to facilitate a flow-visualization study in the future,
and had a cross-section of 76.2 mm in width by 254 mm in
height. These dimensions were carefully selected to ensure
coincidence between the first acoustic resonance frequency (fa = 688 Hz) and the frequency of vortex shedding from
single, tandem, and side-by side cylinders.
Several cylinders with different diameters were used. The
largest diameter produced a maximum wind tunnel blockage ratio of about 10%. The cylinders were made of aluminum and were rigidly mounted on the test section sidewall
to eliminate the effect of cylinder vibration on the flowsound interaction mechanism. Two windows attached to
the sidewalls were used to provide flexibility in setting up
different cylinder diameters, different arrangements and
different spacing ratios.
The flow velocity inside the test section was calibrated by
means of a pitot tube and a pressure transducer. The fluctuating pressure on the top wall of the duct was measured by
means of a 1/4” condenser microphone, which was flushmounted on the test section wall at the location of the maximum acoustic pressure. A Hewlett-Packard analyzer was
used for spectral analysis. Each spectrum was obtained by
averaging 100 samples. Data was collected at a sampling
rate of 4100 Hz. More details about the experimental setup and instrumentation can be found in Mohany and Ziada
[17].
3. Aeroacoustic Response of a Single Cylinder
Figure 2(a) shows a typical pressure spectrum for a 19
mm single cylinder in cross-flow, and Figure 2(b) shows a
waterfall plot of pressure spectra for the same cylinder. At
off-resonance conditions, two frequency components can
be observed in Figure 2(a), the lower component is the vortex shedding frequency, while the higher component near
688 Hz is the first acoustic mode of the duct housing the
cylinder. As the flow velocity is increased, the frequency
of the vortex shedding component and its amplitude are
increased until the vortex shedding frequency becomes
close to that of the acoustic mode, where the lock-in phenomenon is initiated and an intense acoustic resonance
is produced, as shown in Figure 2(b). The recorded maximum sound pressure level for this case is 155.6 dB. As
can be seen in Figure 2(b), during the lock-in range, high
frequency components are also generated. These components are not the higher acoustic modes (f(2λ/2), f(3λ/2),
f(4λ/2),…) but rather the higher harmonics of the first
acoustic mode (2f(λ/2), 3f(λ/2), 4f(λ/2),…). These higher
harmonics are generated by nonlinear effects due to the
very high acoustic pressure at resonance. Figure 3 is constructed from Figure 2(b), and depicts the frequency and
the amplitude of the vortex shedding component in pressure spectra as functions of the reduced velocity. Similar response was obtained for a 25.4 mm single cylinder in crossflow as shown in Figure 4. The lock-in region for this case is
wider and the maximum acoustic pressure at resonance is
AECL NUCLEAR REVIEW
47
self-excited acoustic resonance of isolated cylinders in cross-flow
a. mohany
aecl Nuclear Review
vol 1, Number 1, june 2012
much higher than those observed for all other tested cases
of single cylinders. The reduced velocity Ur presented in all
figures is based on the frequency of the lowest cross-mode
of the test section, fa, as illustrated in equation 1;
Ur =
U
fa ×D
(1)
It is clear from Figures 3 and 4 that within a certain range of
Reynolds number, the Strouhal number for a single cylinder
is about 0.2. If acoustic resonance is initiated at a constant
Strouhal number, given by equation 2, increasing the diameter will also increase the critical flow velocity U for the onset of resonance.
f ×D
(2)
St = a
U
This feature is illustrated in Figure 5, which shows the vortex shedding frequency and the lock-in range for four single
cylinders with different diameters, D = 0.5”, 0.625”, 0.75”,
and 1”. It is clear from this figure that increasing the cylinder diameter will result in higher critical flow velocity and
dynamic head at the onset of resonance. The flow excitation
energy at the resonance conditions will therefore be higher
for larger diameter cylinders.
Figure 2(B)
Pressure spectra measured on the top wall. Single cylinder
tests; D = 19 mm [17].
1400
6
1200
5
Frequency (Hz)
Acoustic pressure (Pa)
vortex shedding
4
3
2
1st acoustic mode
800
fa = 680
600
400
St = 0.201
200
1
0
0
0
600
1200
1800
2400
3000
2
0
2
4
6
8
10
4
6
8
10
1200
Acoustic pressure (Pa)
A numerical simulation of the flow-excited acoustic resonance is performed to reveal the details of flow-sound interaction mechanisms, including the nature and the locations of the aeroacoustic sources in the flow field. A deep
understanding of the flow-sound interaction mechanism
and the resulting energy transfer between the flow and the
resonant sound fields would improve our ability to alleviate
and control the occurrence of acoustic resonance. This can
be achieved by identifying the location of the aeroacoustic
sources in the flow field and developing suitable means to
mitigate the energy transfer mechanism [18]. In this simulation, the flow field and the sound field are simulated
0
1400
Frequency (Hz)
Figure 2(a)
Pressure spectrum measured on the top wall of the test section for a single cylinder with D = 19 mm at a flow velocity
of 25 m/s [17].
48
1000
1000
800
600
400
200
0
Reduced velocity
Figure 3
Frequency and amplitude of pressure fluctuation on the
test section top wall at the frequency of vortex shedding.
Single cylinder tests; D = 19 mm [17].
aecl Nuclear Review
vol 1, Number 1, june 2012
self-excited acoustic resonance of isolated cylinders in cross-flow
a. mohany
1000
fa = 664
600
400
200
St = 0.199
0
0
2
4
6
8
  
Π = − ρ ∫ ω ⋅ (u × v )d∀
Acoustic pressure (Pa)
2500
2000
1500
1000
500
0
0
2
4
6
8
Reduced velocity
Figure 4
Frequency and amplitude of pressure fluctuation on the
test section top wall at the frequency of vortex shedding.
Single cylinder tests; D = 25.4 mm.
1400
D = 0.5''
Frequency (Hz)
1200
D = 0.625"
D = 0.75''
1000
D = 1.0''
800
600
400
12.5
200
0
(3)
Where Π is the instantaneous acoustic power,
is the
vorticity,
is flow velocity, and
is acoustic particle velocity. As mentioned earlier, to account for the effect of the
resonant sound on modifying the flow around the cylinder,
a cross-flow oscillation was applied at the top and the bottom boundaries as well as at the inlet to mimic the acoustic
resonance of the first cross mode. The amplitude and the
frequency of the cross-flow perturbation were changed independently to examine the sensitivity of the flow to this
perturbation and also to determine the minimum amplitude at which the lock-in occurs. Figure 6 shows the dependence of the lock-in on the amplitude and frequency of
the cross-flow oscillation for the case of a single cylinder in
cross-flow. It is found that the minimum perturbation amplitude at which the lock-in occurs was 2.5% of the main
flow velocity. This minimum amplitude was obtained near
the condition of frequency coincidence, i.e., when the excitation frequency, fa, approximated the natural vortex shedding frequency, fv, (0.975< fa/fv <1.025). As the excitation
frequency deviated from the vortex shedding frequency, the
minimum amplitude required to lock-in the shedding frequency increased as shown in Figure 6. The lock-in region
obtained from the simulation is indicated by the broken line
in Figure 6.
0
20
40
60
80
100
120
Velocity (m/s)
Figure 5
Aeroacoustic response of a single cylinder in cross-flow.
separately (i.e., uncoupled). In order to simulate the coupling phenomenon, the simulation of the unsteady flow
field was repeated with applying an oscillatory cross-flow
velocity perturbation, which simulates the acoustic particle velocity of the resonant sound field. The acoustic
sources (or sinks) in the flow field are then predicted by
combining the solution of the acoustic field, which is modeled by the finite-element method using ABAQUS, with
Oscillation amplitude (v/U )%
Frequency (Hz)
800
the acoustically coupled unsteady flow field, which is modeled using the Spalart-Allmaras (S-A) turbulent model in
FLUENT to predict the vorticity distribution in the cylinder
wake. Howe’s theory of aerodynamic sound [19, 20], equation 3, is then used to calculate the instantaneous acoustic
power generated by the convection of the unsteady vorticity field within the sound field. The net acoustic energy is
the integration of the instantaneous acoustic power over a
complete acoustic cycle. Therefore, the locations within the
flow field where the acoustic energy is either absorbed or
generated can be identified. Details of the numerical simulation can be found in [21].
10
Lock-in
7.5
5
No
lock-in
2.5
0
0.75
0.875
1
1.125
1.25
f a /f v
Figure 6
Dependence of the lock-in on the amplitude and frequency
of the cross-flow oscillation for the case of a single cylinder.
♦ , Lock-in; ○, No lock-in [21].
AECL NUCLEAR REVIEW
49
self-excited acoustic resonance of isolated cylinders in cross-flow
a. mohany
aecl Nuclear Review
vol 1, Number 1, june 2012
from that of a single cylinder in cross-flow. Figure 9 represents the frequency and the amplitude of the vortex shedding component in pressure spectra as functions of the reduced velocity for two tandem cylinders with spacing ratio,
L/D = 2. It is clear from Figure 9 that the acoustic resonance
of the first mode occurs over two different ranges of flow velocity. The first resonance range occurs before the velocity
of frequency coincidence and is called the pre-coincidence
acoustic resonance, while the second resonance range occurs at the velocity of frequency coincidence and is called
the coincidence acoustic resonance. More details about
these resonance ranges can be found in references [17, 22].
800
Frequency (Hz)
Figure 7 shows the total acoustic energy over one cycle. The
red colour in these figures represents net positive energy
and the blue colour represents net negative energy. The total
energy transfer per cycle for different downstream locations
is shown in Figure 8. For each streamwise location, the total
energy represents the energy integrated over a rectangular
area with a height extending along y = ± 10 D and a small
width of rx centered at the streamwise location x. As can
be seen from this figure, the total energy transfer just downstream of the cylinder is positive and is larger than that occurring in the downstream region, which means that there
is a strong acoustic source located there, resulting in a net
positive energy transfer from the flow field to the sound field.
600
400
200
0
5.0
4.0
Figure 7
Total acoustic energy over one cycle. fa/fv = 1. [21].
P*
3.0
2.0
1.0
0.0
0
2
4
6
8
10
Reduced velocity
Figure 9
Aeroacoustic response of two tandem cylinders with L/D =
2; D = 25.4 mm.
-2
0
2
4
6
8
10
12
X/D
Figure 8
Total energy transfer per cycle at different downstream locations. Mohany and Ziada [21].
4. Aeroacoustic Response of Two Tandem Cylinders
50
The aeroacoustic response for the case of two tandem cylinders in cross-flow with spacing ratios within the proximity interference region, L/D < 3.5, is considerably different
It should be noted here that the acoustic pressure, Prms, in
Figure 9 is normalized by the dynamic head, ½ ρU2, and the
Mach number, M, as shown in equation 4;
P* 
Prms
1
U 2 M
2
(4)
This normalization procedure has been suggested by
Blevins [23] and has been used by Mohany and Ziada [17]
to properly account for the main parameters influencing
the sound pressure level of acoustic resonances excited by
vortex shedding from single cylinders. As the spacing ratio
between the tandem cylinders is increased beyond 3.5, the
flow pattern changes to that corresponding to the wake interference regime. The aeroacoustic response at these large
spacing ratios is found to be similar to that of a single cylinder. A typical example for L/D = 4.5 is shown in Figure 10.
It is also observed that as the spacing ratio is increased, the
aecl Nuclear Review
vol 1, Number 1, june 2012
self-excited acoustic resonance of isolated cylinders in cross-flow
a. mohany
Strouhal number increases towards 0.2 and the amplitude
of the maximum acoustic pressure decreases, which indicates a transition to the single cylinder features.
A comparison between the pre-coincidence acoustic resonance and the acoustic resonance mechanism of in-line
tube bundles shows interesting similarity. The acoustic resonance of in-line tube bundles normally starts and subsides
before the flow velocity reaches that corresponding to the
frequency coincidence conditions. This similarity is highlighted in Figure 11, which compares the acoustic response
of tandem cylinders with that of in-line tube bundles with
nearly equal streamwise tube spacing ratio. The Strouhal
number of vortex shedding for the in-line tube bundles is
seen to be similar to that for the case of two tandem cylinders. Moreover, the acoustic resonance for in-line tube
bundles and the pre-coincidence resonance of tandem cylinders are seen to be initiated at the same reduced velocity.
With respect to the resonance intensity, it is clearly stronger
for tube bundles because they have lower radiation losses
and larger number of acoustic sources than in the case of
two tandem cylinders.
cylinder diameter, the first acoustic mode is excited over a
single range of flow velocity, which is similar to the resonance mechanism of a single cylinder in cross-flow. Figure
12 shows a typical response of two tandem cylinders with
a spacing ratio of L/D = 2.5, which is the same spacing ratio
as that of Figure 11 but for smaller cylinder diameter, D =
7.6 mm. Figure 13 summarizes the results of the aeroacoustic response for the case of two tandem cylinders in crossflow with different spacing ratios and cylinder diameters.
The solid data points correspond to the cases when the precoincidence resonance is excited. It is clear from Figure 13
that as the spacing ratio increases, the flow energy required
to excite the pre-coincidence acoustic resonance decreases,
which means that the pre-coincidence resonance is easier
to excite for large spacing ratios with smaller diameter cylinders in comparison with that for small spacing ratios.
As discussed in Mohany and Ziada [22], the pre-coincidence acoustic resonance of two tandem cylinders with
spacing ratios within the proximity interference region is
only observed when the cylinder diameter exceeds a certain value. However, for two tandem cylinders with small
Frequency (Hz)
1600
1200
800
400
St = 0.172
0
0
2
4
6
8
10
12
14
2
1.5
P*
Figure 11
Comparison between the acoustic response of two tandem
cylinders with that of in-line tube bundles. •, two tandem
cylinders with L/D = 2.5; p, in-line tube bundles with XL =
2.6 and XT = 3.0 [22].
1
0.5
0
0
2
4
6
8
Reduced velocity
10
12
14
Figure 10
Aeroacoustic response of two tandem cylinders. L/D = 4.5;
D = 12.7mm, Mohany and Ziada [17].
Mohany and Ziada [21] conducted a numerical simulation
of the flow-excited acoustic resonance for the case of two
tandem cylinders in cross-flow. The wake structure behind
the tandem cylinders before applying the cross-flow oscillation is shown in Figure 14. The wake structure around
AECL NUCLEAR REVIEW
51
self-excited acoustic resonance of isolated cylinders in cross-flow
a. mohany
aecl Nuclear Review
vol 1, Number 1, june 2012
Frequency
900
800
700
600
500
400
300
200
100
0
St = 0.148
0
1
2
3
4
5
6
7
8
0.5
0.375
P*
0.25
Figure 14
Vorticity contours behind two tandem cylinders, L/D = 2.5,
at Re = 25000 before applying the cross-flow oscillation.
0.125
0
0
2
4
6
8
Reduced Velocity
Figure 12
Aeroacoustic response of two tandem cylinders with L/D =
2.5; D = 7.6 mm, Mohany and Ziada [22].
the tandem cylinders after applying the cross-flow oscillation at a frequency ratio of 1.2, which corresponds to the
pre-coincidence acoustic resonance range, is shown in Figure 15. A comparison between Figures 15 and 14 shows
that for the pre-coincidence acoustic resonance, the wavelength of the wake structure is smaller than that before applying the cross-flow oscillation. This is because the wake
for this case is locked into an oscillation frequency which is
higher than the vortex shedding frequency, fa = 1.2 fv.
3.5
dual resonance
3
no dual resonance
2.5
L/D
2
1.5
1
0
500
1000
1500
2000
2500
3000
3500
4000
Dynamic head *M
Figure 13
Dependence of the pre-coincidence acoustic resonance on
the spacing ratio and the flow excitation energy.
52
Figure 15
Vorticity contours behind two tandem cylinders, L/D = 2.5,
at Re = 25000, pre-coincidence acoustic resonance, fa = 1.2
fv.
Figure 16 represents the total acoustic energy over one cycle during the pre-coincidence acoustic resonance for L/D
= 2.5 and Figure 17 is the distribution of the total energy
transfer at different downstream locations. It is clear from
Figures 16 and 17 that during the pre-coincidence acoustic
resonance a strong acoustic source in the flow field is located in the gap between the cylinders. Moreover, the energy
transfer from the flow field to the sound field associated
with this acoustic source is dominant. Based on extensive
measurements of the dynamic lift force acting on the tandem cylinders and its phase with respect to the resonant
sound field in combination with flow visualization, Mohany
and Ziada [24] suggested that the pre-coincidence acoustic resonance is caused by the instability of the separated
shear layers in the gap between the cylinders.
5. Aeroacoustic Response of Two Side-by-Side Cylinders
For two side-by-side cylinders with small and intermediate spacing ratios, T/D < 2.2, the wakes exhibit bi-stable
flow features as those reported in the literature [25, 26,
27]. Two Strouhal numbers of vortex shedding are observed before the onset of acoustic resonance. However,
aecl Nuclear Review
vol 1, Number 1, june 2012
self-excited acoustic resonance of isolated cylinders in cross-flow
a. mohany
which has occurred in this situation. From the presented
data, it appears that the resonant sound field synchronizes
the vortex shedding process and thereby eliminates the bistable flow phenomenon. Vortex shedding in this case must
occur at an intermediate frequency between the two frequency components associated with the bi-stable flow.
Figure 16
Total acoustic energy over one cycle for two tandem cylinders at conditions simulating the pre-coincidence acoustic
resonance. L/D = 2.5; Re = 25000; fa/fv = 1.2 [21].
+ve
-ve
-2
0
2
4
x/D
6
8
10
12
Figure 17
Total energy transfer per cycle for different downstream locations. Two tandem cylinders at conditions simulating the
pre-coincidence acoustic resonance. L/D = 2.5; Re = 25000;
fa/fv = 1.2 [21].
during the experiments, self-excited acoustic resonance occurred at an intermediate Strouhal number which lies between those corresponding to the bi-stable flow regimes.
Figure 18 represents the aeroacoustic response of two sideby-side cylinders in cross-flow with spacing ratio of T/D =
1.25. In this case, two vortex shedding frequencies are distinguished at the off resonance conditions, which are generated by the narrow and wide wakes behind the cylinders.
The two Strouhal numbers observed in this case are 0.12
and 0.36, as indicated by the straight lines in Figure 18. It
is seen that the lock-in range does not correspond to either
of these Strouhal numbers. Interestingly, after the lock-in
resonance subsides near Ur = 5.6, the low frequency component of vortex shedding is recovered and its frequency
is still lower than the acoustic resonance frequency. Parker and Stoneman [28] describe this behaviour as “locking
up” for resonance that occurs above a natural frequency,
Figure 18
Aeroacoustic response of two side-by-side cylinders with
T/D = 1.25 and D = 21.8 mm. •, lower vortex shedding frequency; p, higher vortex shedding frequency; ♦, first resonance mode frequency, Hanson et al. (2009).
To confirm this supposition, a numerical simulation of the
flow-sound interaction mechanism for this case was performed. Figure 19 shows the wake structure behind the two
cylinders with spacing ratio of T/D = 1.25 at the off resonance condition, which shows clearly the bi-stable flow regimes. The wake structure behind the cylinders after applying the cross-flow perturbation, which mimic the excitation
of acoustic resonance, is shown in Figure 20. A comparison
between Figures 19 and 20, shows that after applying the
cross-flow perturbation, the flow in the wakes of the cylinders is synchronized and thereby the bi-stable biased flow
is eliminated. More details can be found in Mohany et al.
[30].
The aeroacoustic response of two side-by-side cylinders
with large spacing ratios, T/D > 2.2, exhibits a single vortex shedding frequency of about 0.22, which agrees well
with the values reported in the literature. The acoustic
AECL NUCLEAR REVIEW
53
aecl Nuclear Review
vol 1, Number 1, june 2012
self-excited acoustic resonance of isolated cylinders in cross-flow
a. mohany
Figure 19
Vorticity contours behind two side-by-side cylinders, T/D =
1.25, at Re = 25000, no acoustic excitation.
Figure 20
Vorticity contours behind two side-by-side cylinders, T/D =
1.25, at Re = 25000, with acoustic excitation.
resonance is excited by the vortex shedding observed before
the onset of resonance. This behavior is clear in Figure 21,
which represents the aeroacoustic response of two side-byside cylinders in cross-flow with spacing ratio of T/D = 2.5.
A numerical simulation is performed for this case and the
wake structure behind the cylinders at the off resonance
conditions is shown in Figure 22. It is clear from this figure
that a symmetric wake structure behind the cylinders with
one vortex shedding frequency, which is similar to that of a
single cylinder in cross-flow, is formed.
Figure 21
The aeroacoustic response of the two side-by-side cylinders
with T/D = 2.5 and D = 21.8 mm. •, vortex shedding frequency; ♦, first resonance mode frequency [29].
6. Conclusion
54
A summary of a research program on the acoustic resonance
mechanism of single, two tandem, and two side-by-side
cylinders is presented in this paper. The aeroacoustic response of a single cylinder is similar to that reported in the
literature. Acoustic resonance is excited by the natural vortex shedding observed before the onset of resonance. The
Strouhal number of vortex shedding is 0.2 ± 0.005. As the
diameter for the single cylinder is increased, the required
flow velocity to excite the acoustic resonance increases,
which results in a stronger acoustic resonance and a wider
lock-in range of flow velocity. The acoustic resonance of the
tandem cylinders within the proximity interference region
occurs over two different velocity ranges. The one occurring at higher reduced velocities, the coincidence resonance
Figure 22
Vorticity contours behind two side-by-side cylinders, T/D =
2.5, at Re = 25000, no acoustic excitation.
range, is excited by the natural vortex shedding process, as
in the single cylinder case. The other range, the pre-coincidence resonance range, is initiated at a much lower reduced
velocity, and appears to be excited by the shear layers that
aecl Nuclear Review
vol 1, Number 1, june 2012
self-excited acoustic resonance of isolated cylinders in cross-flow
a. mohany
form in the gap between the two cylinders. Moreover, a comparison between the pre-coincidence acoustic resonance
and the acoustic resonance mechanism of in-line tube bundles shows interesting similarity. For two side-by-side cylinders with small and intermediate spacing ratios, i.e., T/D
< 2.2, the flow is shown to exhibit a bi-stable behaviour at
the off resonance conditions. Acoustic resonance for this
case is excited at an intermediate Strouhal number, which lies
between those corresponding to the bi-stable flow regimes.
However, for two side-by-side cylinders with large spacing
ratios, T/D > 2.2, the flow at the off resonance conditions is
symmetric and exhibit a single vortex shedding frequency.
The acoustic resonance for this case is excited by the vortex
shedding observed before the onset of resonance.
References
[1] M.M. Zdravkovich, 1977, “Review of Flow Interference Between Two Circular Cylinders in Various Arrangements”. Transactions of the ASME, Journal of Fluids Engineering, 99(4), pp. 618-633.
[2] M.M. Zdravkovich, 1985, “Flow Induced Oscillations of Two Interfering Circular Cylinders”. Journal of Sound and Vibration, 101(4), pp. 511-521
[3] T. Igarashi, 1981, “Characteristics of the Flow Around Two Circular Cylinders Arranged in Tandem - 1st report”. Bulletin of the Japan Society of Mechanical Engineers (JSME), 24(188), pp. 323331
[4] T. Igarashi, 1984, “Characteristics of the Flow Around Two Circular Cylinders Arranged in Tandem: 2nd Report”, Unique phenomenon at small spacing. Bulletin of the Japan Society of Mechanical Engineers (JSME), 27(233), pp. 2380-2387
[5] L. Ljungkrona, Ch. Norberg, and B. Sunden, 1991, “Free-Stream Turbulence and Tube Spacing Effects on Surface Pressure Fluctuations for Two Tubes in an In-Line Arrangement”. Journal of
Fluids and Structures, 5(6), pp.701-727
[6] J.-C. Lin, Y. Yang, and D. Rockwell, 2002, “Flow Past Two Cylinders in Tandem: Instantaneous and Averaged Flow Structure”. Journal of Fluids and Structures, 16(8), pp. 1059-1071
[7] Md.M. Alam, M. Moriya, K. Takai, and H. Sakamoto, 2003, “Fluctuating Fluid Forces Acting on Two Circular Cylinders in a Tandem Arrangement at a Subcritical Reynolds Number”. Journal of
Wind Engineering and Industrial Aerodynamics 91(1-2), pp. 139-154
[8] D. Sumner, M.D. Richards, and O.O. Akosile, 2008, “Strouhal Number Data for Two Staggered Circular Cylinders”. Journal of Wind Engineering and Industrial Aerodynamics 96(6-7), pp. 859-871
[9] D. Sumner, 2010, “Two Circular Cylinders in Cross-flow: A Review”. Journal of Fluids and Structures 26(6), pp. 849-899
[10] R.D. Blevins, 1985, “The Effect of Sound on Vortex Shedding from Cylinders”. Journal of Fluid Mechanics, 161, pp. 217-237.
[11] P. Anagnostopoulos, 2002, “Flow-Induced Vibrations in Engineering Practice”, WIT Press
[12] G.H. Koopmann, 1967, The Vortex Wakes of Vibrating Cylinders at Low Reynolds Numbers. Journal of Fluid Mechanics, 28(3), pp. 501-512
[13] O.M. Griffin, and S.E. Ramberg, 1974, “The Vortex-street Wakes of Vibrating Cylinders”. Journal of Fluid Mechanics, 66(3), pp. 553-576
[14] R.E.D. Bishop, and A.Y. Hassan, 1964, “The Lift and Drag Forces on a Circular Cylinder Oscillating in a Flowing Fluid”. Proceedings of the Royal Society of London, Series A, 277(1368), pp. 51-75
[15] M.M. Zdravkovich, 1982, “Modification of Vortex Shedding in the Synchronization Range”, Transactions of the ASME, Journal of Fluids Engineering, 104(4), pp.513-517
[16] R.D. Blevins and M.M Bressler, 1993, “Experiments on Acoustic Resonance in Heat Exchanger Tube Bundles”. Journal of Sound and Vibration, 164(3), pp. 503-533.
[17] A. Mohany, and S. Ziada, 2005, “Flow-excited Acoustic Resonance of Two Tandem Cylinders in Cross-flow”. Journal of Fluids and Structures, 21(1), pp. 103-119
[18] S.A.T. Stoneman, K. Hourigan, A.N. Stokes, and M.C. Welsh, 1988, “Resonant Sound Caused by Flow Past Two Plates in Tandem in a Duct”, Journal of Fluid Mechanics, 192, pp. 455-484
[19] M.S. Howe, 1975, “Contributions to the Theory of Aerodynamic Sound, with Application to Excess Jet Noise and the Theory of the Flute”. Journal of Fluid Mechanics, 71(4), pp. 625-673
[20] M.S. Howe, 1984, “On the Absorption of Sound by Turbulence and Other Hydrodynamic Flows”. IMA Journal of Applied Mathematics, 32(1-3), pp. 187-209
[21] A. Mohany, and S. Ziada, 2009, “Numerical Simulation of the Flow-Sound Interaction Mechanisms of a Single and Two-Tandem Cylinders in Cross-Flow”. Journal of Pressure Vessel Technology,
131(3), p. 031306
[22] A. Mohany, and S. Ziada, 2009, “A Parametric Study of the Resonance Mechanism of Two Tandem Cylinders in Cross-flow”. Journal of Pressure Vessel Technology, 131(2), p. 021302
[23] R.D. Blevins, 1990, “Flow-Induced Vibration”, 2nd ed., Van Nostrand Reinhold, New York.
[24] A. Mohany, and S. Ziada, 2009, “Effect of Acoustic Resonance on the Dynamic Lift Forces Acting on Two Tandem Cylinders in Cross-flow”. Journal of Fluids and Structures, 25(3), pp. 461-478.
[25] P.W. Bearman, and A.J. Wadcock, 1973, “The interaction Between a Pair of Circular Cylinders Normal to a Stream”. Journal of Fluid Mechanics, 61(3), pp. 499-511
[26] H.J. Kim, and P.A. Durbin, 1988, “Investigation of the Flow Between a Pair of Circular Cylinders in the Flopping Regime”. Journal of Fluid Mechanics, 196, pp. 431-448.
[27] D. Sumner, S.S.T. Wong, S.J. Price, and M.P. Paidoussis, 1999, “Fluid Behaviour of Side-by-side Circular Cylinders in Steady Cross-flow”. Journal of Fluids and Structures, 13(3), pp. 309-338
[28] R. Parker, and S.A.T. Stoneman, 1989, “The Excitation and Consequences of Acoustic Resonances in Enclosed Fluid Flow around Solid Bodies”. Proceedings of the Institution of Mechanical
Engineers, Part C: Journal of Mechanical and Engineering Science, 203(1), pp. 9-19
[29] R. Hanson, A. Mohany, and S. Ziada, 2009, “Flow-excited Acoustic Resonance of Two Side-by-side Cylinders in Cross-flow”. Journal of Fluids and Structures, 25(1), pp. 80-94.
[30] A. Mohany, M. Hassam, and S. Ziada, 2011, “Numerical Simulation of the Flow-sound Interaction Mechanisms of Two Side-by-side Cylinders in Cross-flow”. ASME Pressure Vessels & Piping
Conference (PVP 2011), July 17-21, Baltimore, Maryland, USA, PVP2011-57282
AECL NUCLEAR REVIEW
55
56
TECHNICAL NOTE
Abstract
A boron-loaded liquid scintillator (LS)
has been optimized for neutron detection
application in a high gamma field
environment. It is composed of the solvent
linear alkylbenzene (LAB), a boroncontaining material, o-carborane (C2B10H12);
a fluor, 2,5-diphenyloxazole (PPO); and a
wavelength shifter, 1,4-bis[2-methylstyryl]
benzene (bis-MSB). Preparation of the liquid
a novel boron-loaded liquid
scintillator for neutron detection
G. Bentoumi*, X. Dai, E. Pruszkowski, L. Li, B. Sur
Atomic Energy of Canada Limited, Chalk River Laboratories, Chalk River, Ontario, Canada, K0J 1J0
Article Info
Keywords: scintillation, boron, linear alkylbenzene, neutron
Article history: Received 13 April 2012, Accepted 25 June 2012, Available online 30 June 2012.
*Corresponding Author: (613) 584-3311 ext. 46727, [email protected]
Abbreviations
LS - Liquid scintillator; LAB - Linear alkylbenzene; PPO - 2,5-diphenyloxazole; bis-MSB - 1,4-bis[2-methylstyryl]benzene; NRU - National Research Universal reactor; PMT - Photomultiplier tube; UV - Ultraviolet; IR - Infrared
scintillator and optimization of its chemical
composition are described. The boronloaded LS has been tested with a neutron
beam at the National Research Universal
(NRU) reactor. A peak at an equivalent
energy of 60 keV is observed in the energy
spectrum and is attributed to neutrons. The
results confirm the possibility of using B-10
loaded scintillator as a sensitive medium
for neutron detection in a relatively large
background of gamma rays.
1. Introduction
Helium-3 filled ionization chamber tubes have been extensively used in neutron
detection because of their good neutron-to-gamma discrimination capability. However, there is currently a global shortage of 3He. As a result, it is necessary to design
new neutron detectors relying on other mechanisms that are as effective as the 3He
detectors. Thermal neutron capture by boron-10 (10B) has a cross section of 3838
barns, which leads to an average deposit of 2.34 MeV in the surrounding medium
[1]. In a scintillator, part of this energy transforms to optical photons that can be
detected easily using a photomultiplier tube (PMT) [1, 2]. Neutron detectors using
solid, liquid or gas scintillators have been studied for replacing 3He detectors.
Several 10B-loaded liquid scintillator compositions are commercially available.
However, they have several drawbacks. For instance, they typically contain a high
content of Trimethyl Borate, which is unstable and decomposes in contact with
moisture. These liquid scintillators are also flammable and yield lower light output.
Therefore, new and improved liquid scintillators are needed for effective detection
of neutrons.
2. Method
A general-purpose liquid scintillator based on the solvent linear alkyl benzene (LAB) is
proposed herein. It has the general formula C6H5CnH2n+1, where n is an integer between
10 and 16. LAB is produced easily and has a high flash point (130 °C) [3]. Consequently
a low-cost, high efficient and safe neutron detector can be prepared. In general, scintillator light output is enhanced by adding fluor PPO and wavelength-shifter bis-MSB.
Due to the nature of charge transfer between the solvent (LAB) and solutes (PPO
and bis-MSB) and the light scintillation process, the optimization of the solutes’
concentrations is of great importance. For this, we initially varied the PPO concentration from 0 to 30 g/l. When optimal concentration of PPO was established, the
concentration of bis-MSB was varied from 0 to 1g/l. Solutions were produced by
combining in different proportion pure LAB and highly concentrated LAB solution
with 30 g/L PPO and 0.2 g/L bis-MSB. For light yield quantification, the samples
were transferred to a 10 mm cylindrical quartz cell and a UV light source was used
for excitation. Emitted blue light, primarily around 425 nm, was collected by an
optical fibre for analysis with an Ocean Optics HR4000 UV-NIR spectrometer. Figure 1 shows an example of the emission spectrum for the scintillator LAB with the
solutes PPO and bisMSB. A suitable integration time was chosen to allow accurate
readouts. The area underneath the emission peak was determined. Firstly, each
sample’s yield curve was divided by its respective integration time to normalize
AECL NUCLEAR REVIEW
57
a novel boron-loaded liquid scintillator for neutron detection
g. bentoumi, x. dai, e. pruszkowski, l. li and b. sur
aecl Nuclear Review
vol 1, Number 1, june 2012
80
Total light yield (Arb. Unit)
Light yield (Arb. unit)
1.0
0.8
0.6
0.4
75
70
65
0.2
0
400
600
800
400
600
800
1000
Concentration of bis-MSB (mg/L)
Wavelength (nm)
Figure 1
Emission spectra of scintillator including solvent LAB and
solutes PPO and bisMSB.
Figure 3
Light yield variation versus 1,4-bis[2-methylstyryl]benzene
(bis-MSB) concentration in LAB.
the results. Then, the background light corresponding to
range 600 - 650 nm was subtracted from the curve. Lastly,
the area under each peak was determined using a Riemann
sum.
is somewhat similar by increasing the bis-MSB concentration. Thus, the optimal concentration for maximum light
yield is 5 g/L of PPO and nearly 500 mg/L of bis-MSB.
3. Results and Conclusion
Total light yield (Arb. Unit)
The results, as shown on Figures 2 and 3, show that scintillator light yield depends strongly on the concentrations of
the fluors. For low PPO concentration (< 5g/l), light yield
increases rapidly before saturation. For PPO concentration more than 5g/l, light yield is constant. The situation
60
40
20
0
5
10
15
20
25
30
Concentration of PPO (g/L)
Figure 2
Light yield variation versus 2,5-diphenyloxazole (PPO) concentration in LAB.
58
200
1000
In order to test performance of this optimized liquid
scintillator, 5% of O-carborane (C2H12B10) containing natural boron was loaded in the LS. As stated in the equation:
B neutron capture leads to creation of an alpha particle
with an average energy of 1.48 MeV. It was demonstrated
early [4,5] that in liquid scintillator, the quenching factor for
1 to 2 MeV alpha particles is in the range 20-25. Hence 10B
neutron capture is expected to generate a peak in the range
60-75 keV.
10
As shown on Figure 4, tests with a neutron beam were done
at the National Research Universal (NRU) reactor. Due to
the presence of gamma radiation and in order to extract
the neutron contribution in the detected signal, four types
of measurements were performed using combinations of
cadmium sheet (to block neutron beam) and lead bricks (to
stop gamma rays). The energy spectrum was acquired in
four different conditions and calibrated using a Cobalt-60
source. Figure 5 shows the four energy spectra recorded for
about 5 minutes. The presence of a peak around 60 keV is
directly correlated to neutrons. Based on that correlation
and the excepted equivalent energy for such capture interaction [2, 4, 5], the peak was definitely attributed to neutron
capture by 10B. To our knowledge, this is the first time such a
measurement has been made with a LAB-based scintillator.
aecl Nuclear Review
vol 1, Number 1, june 2012
a novel boron-loaded liquid scintillator for neutron detection
g. bentoumi, x. dai, e. pruszkowski, l. li and b. sur
DAQ Detector neutron
neutron
neutron
neutron
Scintillator
PMT
Voltage
Divider
Pre­ Ampl
Ampli Oscilloscope Neutron beam
ADC Computer Figure 4
Experimental setup for scintillator characterization.
These measurements were possible despite the presence of
gamma rays and are very encouraging for the use of boronloaded LAB based scintillator in neutrons detection.
In conclusion, a boron-loaded liquid scintillator based on
LAB solvent was prepared and its chemical composition has
been optimized for high light output. Neutron beam measurements have shown a new peak around 60 keV. With a
systematic study, this peak is confirmed to be due to neutron capture by boron-10. This confirmed the feasibility of
using boron-loaded LAB as a safe, effective liquid scintillator for neutron detection.
Intensity (Arb. Unit)
HV
ON, gamma ON
ON, gamma OFF
OFF, gamma OFF
OFF, gamma ON
40
60
Energy (keV)
80
100
Figure 5
Energy spectra recorded in four conditions by the boron
loaded LAB based scintillator. (r) Neutron OFF, gamma
OFF, (•) Neutron ON, Gamma ON, (p) Neutron OFF, Gamma
ON, (◊) Neutron ON, Gamma OFF.
References
[1] Glenn F. Knoll, 2010, “Radiation Detection and Measurement”, 4th ed. USA: John Wiley &Son Ltd.
[2] M. Yeh, A. Garnov, and R.L. Hahn, 2007, “Gadolinium-loaded Liquid Scintillator for High-precision Measurements of Antineutrino Oscillations and the Mixing Angle, θ13”. Nuclear instruments
and Methods in physics Research A, 578(1), pp. 329-339
[3] T. Marrodán Undagoitia, F. von Feilitzsch, L. Oberauer, W. Potzel, and A. Ulrich, J. Winter, and M. Wurm, 2009, “Fluorescence Decay-time Constants in Organic Liquid Scintillators”. Review of
Scientific Instruments, 80(4), paper 043301 (2009)
[4] Wuon-Shik Kim, Hyeon-Soo Kim, Ki-Hwan Kim, Yong-Uhn Kim, and Ki-Hyon Kim, 1997, “Gamma-ray Pulse Height Spectrum of 241Am-Be Source by ‘Li-BC501 (n-γ) Spectrometer System”. Journal of Radioanalytical and Nuclear Chemistry, 215(2), pp. 257-261
[5] H. M. O’Keeffe, E. O’Sullivan, M. C. Chen, 2011, “Scintillation Decay Time and Pulse Shape Discrimination in Oxygenated and Deoxygenated Solutions of Linear Alkylbenzene for the SNO+ Experiment”. Nuclear Instruments and Methods in Physics Research A, 640(1), pp. 119-122
AECL NUCLEAR REVIEW
59
60
TECHNICAL NOTE
Abstract
A radiation imaging system has been
developed using the concept of inverse
collimation, where a narrow shielding pencil
is used instead of a classical collimator.
This imaging detector is smaller, lighter
and less expensive than a traditionally
collimated detector, and can produce a
spherical raster image of radiation sources
in its surroundings. A prototype was
developed at Atomic Energy of Canada
Limited – Chalk River Laboratories, and
the concept has been successfully proven in
experiments using a point source as well as
real sources in a high ambient field area.
Such a radiation imaging system is effective
in locating radiation sources in areas where
accessibility is low and risk of radiological
contamination is high, with applications
in decontamination and decommissioning
activities, nuclear material processing labs,
etc.
inverse collimator-based radiation
imaging detector system
A. Das*, B. Sur, S. Yue, G. Jonkmans
Atomic Energy of Canada Limited, Chalk River Laboratories, Chalk River, Ontario, Canada, K0J 1J0
Article Info
Article history: Received 29 May 2012, Accepted 20 June 2012, Available online 30 June 2012.
*Corresponding Author: (613) 584-3311 ext.43468, [email protected]
1. Introduction
Imaging and visual representation of radiation sources and radiological contamination have applications in several fields: radiation protection, decommissioning
and cleanup, waste management, to name a few. Improvements in techniques for
imaging high radiation fields are making such imaging systems smaller, faster and
more cost effective.
Development of a radiation source imaging system at Chalk River Laboratories
(CRL) was motivated in part by a need to image the sources of radiation inside
a radioactive isotope processing hot-cell. Manual access inside the hot-cell is restricted for radiological safety reasons and equipment entering the cell has size
and weight restrictions. In imaging of radiation sources using non-directional sensors, directionality is achieved through use of heavy shielding collimators. The use
of collimators, including variations such as pinhole and parallel hole collimators,
is a common technique in many prior applications [1, 2, 3, 4] for imaging radiation sources. However, the bulk and weight of this general collimator design often
requires a robust assembly and relatively powerful actuators to manoeuvre the collimated detector, which is expensive in terms of the material and actuators. There
exists a need in many applications for an inexpensive light-weight radiation imaging system. This paper describes a novel approach to the collimator design for a
lighter and economical radiation detector system, developed at CRL.
2. Materials and Methods
The sensor of choice for this application is a silicon PIN photodiode. Si diodes are
simple, robust, low-cost, and have been widely used for measuring gamma dose
rates. The total charge generated in a Si diode is well known to be proportional to
the ionization energy deposited in the diode depletion region, and thus the radiation dose [5]. Consequently, the current in an unbiased Si diode is a measure of the
radiation field or dose rate. In current generation mode, p-n junction diodes as well
as PIN photodiodes have been successfully used as high radiation field detectors
in many facilities at CRL [6, 7, 8]. The photodiode sensor used in this application is
sensitive to gamma dose rates in the range of 10-3 Gy.h-1 to 103 Gy.h-1.
A collimator enables a non-directional sensor to be used in a directional detector system, usually by surrounding the sensor with dense shielding material with
a small aperture, such that radiation from all directions except the aperture is
blocked. In a typical collimator, the detector response is high when the aperture
faces the direction of a source and low elsewhere. An inverse collimator instead
comprises a shielding ‘pencil’ – a thin rod or cone, of dense material that blocks radiation from a narrow solid angle. The detector response is relatively low when the
shielding pencil is pointed towards a strong radiation source, and high otherwise.
This concept is illustrated in Figure 1 for the case of a single point source. The concept of inverse collimation exists in literature [9], albeit purely for planar imaging
of relatively low radiation sources as applicable to nuclear medicine imaging.
AECL NUCLEAR REVIEW
61
aecl Nuclear Review
vol 1, Number 1, june 2012
inversed collimator based radiation imaging detector system
a. das, b. sur, s. yue and g. jonkmans
Figure 1
Collimator vs. inverse collimator, and their ideal response
functions
To realize the inverse collimator concept in an imaging system, a thin rod of gamma blocking material needs to be assembled with the sensor and integrated with a mechanism
that allows the detector and inverse collimator assembly to
be pointed in all directions. A suitable pan-tilt solution was
not found commercially; therefore, an inexpensive module
to pan and tilt the detector and inverse collimator assembly
was designed and built at CRL.
The pan-tilt module employs two stepper motors; one at
the base for rotation in the horizontal plane (pan functionality) and one higher up on the body for rotation in
the vertical plane (tilt function). A thin lead (Pb) pencil
serves as the inverse collimator material in this design;
see Figure 2. The sensor and the inverse collimator pencil are mounted in a diametric spoke of the vertical gear
wheel (175 mm in diameter), which is driven by the tilt
motor through a driver gear. The gear wheel, driver gear
and tilt motor are mounted on a wheel base, which is rotated on the horizontal plane by the pan motor [10]. A
low-noise signal cable attaches to the sensor at its mounting location, with sufficient slack-length to prevent cable
wind-up; a separate power connection is not required.
The entire body of the pan-tilt device is constructed out of
high-performance composite material using a 3D printer.
This provided an inexpensive and radiologically unobtrusive
body, especially for the tilt mechanism that houses the sensor.
This complete imaging system has been patented by AECL.
3. Results and Discussion
62
The imaging detector system was tested using a 37 GBq
(10 Ci) 60Co providing a conical gamma beam. The source
was placed roughly 380 mm from the center of the imaging
Figure 2
Light yield variation versus 2,5-diphenyloxazole (PPO) concentration in LAB.
system. This test served as a proof-of-concept, and allowed measurement of the detector’s response to a simple
source configuration to analyse the accuracy and limitations of the system. The image generated from this test is
presented in Figure 3, as a plot of the gamma field intensities versus directions, represented on a unit sphere. Note
that the numbers on the scale bar are only to be used as
a relative measure. The utility of such an image is to visually present the directions of all sources that contribute to
the radiation field at the detector location, and to identify
the source with the highest contribution. It must be noted
that the raw data collected by the system represents the inverse (or photo-negative) of the desired image; thus a photo-negation algorithm is applied to obtain the image [10].
Figure 3 (a) shows the location of the source as represented
on the sphere. The high signal area on the side away from
the source [Figure 3 (b)], is a result of the sensor’s directionality. The photodiode sensor is rectangular prism shaped;
its sensitivity to gamma rays incident on the tips is roughly
60% that of the sensitivity along the sides of the sensor. The
detector system is configured such that the sensor is longitudinally in line with the inverse collimator, with the tips
facing towards and directly away from the inverse collimator. As a result, a false low measurement occurs when the
aecl Nuclear Review
vol 1, Number 1, june 2012
inversed collimator based radiation imaging detector system
a. das, b. sur, s. yue and g. jonkmans
Figure 3
Gamma radiation source distribution image of a point source (a) side facing source (b) side away from source.
tip of the sensor is facing the source, which translates to a
peak upon image inversion. Thus, for each strong source,
there will be a secondary area of high signal directly opposite to the direction of the source. This effect proves to
be useful in distinguishing erroneous signals and verifying
true source locations.
A practical and more realistic test was performed in order
to test the ability of the system to image two highly radioactive items in the presence of a high background field. A section of pressure tube removed from a CANDU® reactor, and
an irradiated CANDU® fuel pin were placed in a radioactive
materials handling hot-cell, operated by the Materials and
Mechanics branch at CRL. A Si diode gamma detector calibrated in a Co-60 gamma cell [7] was used to measure the
near contact gamma radiation fields for these objects; they
were found to be roughly 1.3 Gy.h-1 for the pressure tube
section and 33 Gy.h-1 for the fuel pin.
The roughly 5 m wide by 3 m deep hot-cell also contained
miscellaneous pieces of equipment and tools that contributed to the ambient gamma field. The detector system was
placed to one side of the cell and the two items used for
this study were placed around it, as depicted in Figure 4.
The fuel pin was laid flat on the floor, roughly 0.3 m from
the base of the detector, and the pressure tube section was
placed vertically around 1 m from the detector.
The image obtained of this setup is shown in Figure 5, with
the outlines of the sources projected onto the surface of the
sphere. The high signal area observed in Figure 5 (a) is in
Figure 4
Multi-source image generation study setup inside hot-cell
the direction expected for the fuel pin. The secondary high
signal area due to the fuel pin can be observed in Figure 5
(b), directly opposite to the fuel pin location. As expected,
the radiation fields observed from the fuel pin were much
higher than the fields from the piece of pressure tube. The
radiation exposure at the detector due to the pressure tube
piece was too small to be resolved in the presence of the
much higher radiation field of the fuel pin.
4. Conclusion
Imaging of radiological environments can be performed
using a directional sensor, where directionality is usually
achieved through heavy collimation. The concept of using
AECL NUCLEAR REVIEW
63
aecl Nuclear Review
vol 1, Number 1, june 2012
inversed collimator based radiation imaging detector system
a. das, b. sur, s. yue and g. jonkmans
Figure 5
Radiation image generated in the hot-cell with the outline of the two sources overlaid as projections onto the sphere. (a) Bright
spot corresponding to the fuel pin, along with superimposed outlines of pressure tube and fuel pin (b) Secondary bright spot
associated with fuel pin, along with pressure tube outline
an inverse collimator, consisting of shielding material in a
narrow solid angle where a typical collimator would have
an aperture, has been proven at CRL. An inverse collimator
is lighter and less expensive, both in terms of material cost
and cost of actuators.
A radiation imaging detector system was designed and built
in-house at CRL, using a silicon photodiode as the gamma
sensor and a lead (Pb) pencil as the shielding material, assembled in a 3D printed composite body. The image is generated by rotating the sensor assembly, performing exposure rate measurements across the 4π solid angle around
the sensor, and assembling the data as a spherical raster
image. The raw image is then inverted to correct the photonegative effect due to the inverse collimator. This system
was demonstrated successfully in a controlled environment
using a 60Co point source.
A test was conducted to determine the ability of the detector system to image two highly radioactive materials in the
presence of high radiation background. The near contact
gamma fields for the two items differed by a factor of 25,
and the less radioactive item was placed farther from the
64
detector. The system was able to detect and image the more
radioactive item, but not the less radioactive item. It is concluded that changes in radiation exposure at the detector as
the inverse collimator was swept through the direction of
the weaker source were too small to be resolved in the presence of the much higher radiation field due to the stronger
source. The test showed a limitation of inverse collimator
systems in detecting radiation sources in the presence of
much stronger sources. Further work is required to better
define and improve the detection thresholds of the system
so that weaker sources may be successfully imaged.
5. Acknowledgements
The authors wish to thank Elzbieta Rochon, Heather Chaput, Joseph Bida and Kevin McCarthy from AECL CRL, for
providing use of their facilities and their assistance in sensor study and testing of the imaging system. The authors
would also like to acknowledge the contributions of Alexandar Mechev and Hinkel Yeung, undergraduate students
from the University of Waterloo, in post-processing for image reconstruction and mechanical design of the hardware,
respectively.
aecl Nuclear Review
vol 1, Number 1, june 2012
References
inversed collimator based radiation imaging detector system
a. das, b. sur, s. yue and g. jonkmans
[1] R. Redus, et. al., October 1995 “An Imaging Nuclear Survey System,” Nuclear Science Symposium and Medical Imaging Conference Record, IEEE, 1, pp. 649-652
[2] W. Lee, G. Cho, 2002, “Pinhole Collimator Design for Nuclear Survey System.” Annals of Nuclear Energy, 29(17), pp. 2029-2040
[3] A.N. Sudarkin, O.P. Ivanov, V.E. Stepanov, A.G. Volkovich, A.S. Turin, A.S. Danilovich, D.D. Rybakov, L.I.N. Urutskoev, 1996, “High-energy Radiation Visualizer (HERV): a New System for Imaging in
X-ray and Ramma-ray Emission Regions.” Recom Ltd., Kurchatov (I.V.) Inst. of Atomic Energy, Moscow, IEEE Transactions on Nuclear Science, 43(4), part 2, pp. 2427-2433
[4] M. Woodring, D. Souza, S. Tipnis, P. Waer, M. Squillante, G. Entine, K.P. Ziock, 1999, “Advanced Radiation Imaging of Low-intensity Gamma-ray Sources.” Nuclear Instruments and Methods in
Physics Research, Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 422(1-3), pp. 709-712
[5] Knoll, G.F., 1989, “Radiation Detection and Measurement,” 2nd Edition, John Wiley & Sons
[6] B. Sur, S. Yue, and A. Thekkevarriam, June 2007, “Radiation Exposure Rate and Liquid Level Measurement Inside a High Level Liquid Waste (HLLW) Storage Tank,” Proceedings of the 28th Annual Conference of the Canadian Nuclear Society, Saint John, New Brunswick, Canada
[7] B. Sur, S. Yue, G. Jonkmans, “A Detector System for Measuring High Radiation Fields,” April 2009, 6th American Nuclear Society International Topical Meeting On Nuclear Plant Instrumentation,
Control, And Human-Machine Interface Technologies (NPIC & HMIT), Knoxville, Tennessee, USA
[8] A. Das, S. Yue, B. Sur, et al., May 2010 “Gamma Radiation Scanning of Nuclear Waste Storage Tile Holes,” Proceedings of the 31st Annual Conference of the Canadian Nuclear Society, Montréal,
Québec, Canada.
[9] D. J. Wagenaar et. al., July 2007 “Inverse Collimation for Nuclear Medicine Imaging,” US Patent 7242003
[10] A. Das, B. Sur, S. Yue, G. Jonkmans, “Detector System for Radiation Imaging Using Inverse Collimation,” June 2011, Proceedings of the 32nd Annual Conference of the Canadian Nuclear Society,
Niagara Falls, Ontario, Canada
66
TECHNICAL NOTE
Abstract
The Ottawa River has received nuclear
reactor effluent from Chalk River
Laboratories (CRL) for more than 60 years,
including releases from a NRX accident
in 1952. Recent interest in the potential
impact of these historical releases and the
possible need for remediation of a small
region immediately downstream from the
release point has led to comprehensive
studies to assess risk to people and
wildlife. In this paper, the results of an
extensive survey of gamma-emitting
anthropogenic radionuclides in Ottawa
River sediment in the vicinity of CRL are
presented. Anthropogenic radionuclides
detected in Ottawa River sediment include
60
Co, 94Nb, 137Cs, 152Eu, 154Eu, 155Eu and
241
Am. Concentrations of all anthropogenic
radionuclides decline rapidly with distance
downstream of the process outfall,
reaching stable concentrations about 2 km
downstream. All of these radionuclides are
found at some sites within 2 km upstream
of the process outfall suggesting limited
upstream transport and sedimentation.
Comparison of anthropogenic radionuclides
with several representative primordial
radionuclides shows that with the exception
of sites at the process outfall and within
2 km downstream of the process outfall,
anthropogenic radionuclides in
Ottawa River sediment near Chalk
River Laboratories
D.J. Rowan*
Atomic Energy of Canada Limited, Chalk River Laboratories, Chalk River, Ontario, Canada, K0J 1J0
Article Info
Article history: Received 19 May 2012, Accepted 26 June 2012, Available online 30 June 2012.
*Corresponding Author: (613) 584-3311 ext. 44732, [email protected]
1. Introduction
The Ottawa River has received nuclear reactor effluent from Chalk River Laboratories (CRL) for more than 60 years, including releases from a NRX accident in 1952.
The process outfall releases liquid effluent through a vertical diffuser located at a
depth of approximately 22 m. In 1992, a “once through” research reactor was permanently shut down, significantly reducing the quantity of fission and activation
products released to the river. Continuing releases are extremely low, approaching or exceeding analytical capabilities. Recent interest in the potential impact of
these historical releases and the possible need for remediation of a small region
immediately downstream from the release point has led to comprehensive studies
to assess risk to people and wildlife. The river at the outfall is public and boaters,
fishermen and campers regularly use this reach of the river.
The aquatic ecosystem at the site has been studied previously, including 137Cs bioaccumulation and biokinetics of invertebrates and fish [1-4] and bioavailability of radionuclides to aquatic macrophytes and associated epiphytes [5]. However, there
has not been a comprehensive study of the benthic community and sediment since
the early 1950’s [6].
In this study, I compare concentrations of anthropogenic radionuclides upstream of
the process outfall are compared, on a transect containing the process outfall, sites
within 2 km downstream of the process outfall referred to as the footprint, and
sites further downstream. Anthropogenic radionuclide concentrations with those
of several primordial radionuclides are also compared.
2. Methods
primordial radionuclide concentrations
greatly exceed CRL derived anthropogenic
radionuclide concentrations. Thus, over
60 years of radionuclide releases from
operations at CRL have had little impact on
radionuclide concentrations in Ottawa River
sediment, except at a few sites immediately
adjacent to the process outfall.
Ottawa River sediment was sampled on transects spaced at 1 km intervals, for distances of about 8 km upstream and downstream of the Chalk River Laboratories
process outfall. Sample sites were chosen at depth internals of 0-5, 5-10, 10-15, 1520, 25-30 and 30-50 m from both shorelines. A total of 214 sites were sampled using a 9” x 9” Ekman dredge. A core sample of the upper 5 cm of sediment was taken
from each Ekman dredge and retained and dried at 0°C. The sediment was ground
and packed in 50 ml vials for radionuclide determination by gamma spectroscopy
using a low-background high purity germanium gamma spectrometer.
3. Results
3.1 Anthropogenic Radionuclides
Cs was detected in all sediment samples (see Table 1). 137Cs in sediments upstream of CRL is due to atmospheric nuclear weapon tests during the early 1960’s,
although the higher concentrations within 2 km upstream of CRL suggests some
137
AECL NUCLEAR REVIEW
67
anthropogenic radionuclides in ottawa river sediment near
chalk river laboratories - d.j. rowan
aecl Nuclear Review
vol 1, Number 1, june 2012
upstream transport of reactor derived 137Cs (see Figure 1).
Concentrations are greatest at the process outfall, rapidly
dropping off within 2 km downstream, to rather stable concentrations from 2-9 km downstream (see Figure 1). Downstream concentrations are about two-fold greater than upstream concentrations (see Table 1), suggesting that CRL
releases have been equivalent to weapon test fallout.
Table 1
Concentrations of anthropogenic radionuclides in sediment
up to 8 km upstream from the process outfall, at the process outfall, up to 2 km downstream of the process outfall,
in the footprint and 2 to 8 km downstream of the process
outfall.
Radionuclide
60
Co (Bq kg-1)
n
94
-1
Nb (Bq kg )
n
137
-1
Cs (Bq kg )
n
152
-1
Eu (Bq kg )
n
154
Eu (Bq kg )
2944±2364
296±216
19±1.6
21
5
35
91
6.5±4.9
18±9.0
12±7.0
1.6±0.5
2
5
26
38
85±5.3
3411±1750
585±145
198±12
83
5
35
91
1.1±0.1
240±171
46±22
3.0±0.2
2
5
22
59
179±134
24±13
1.2±0.1
5
18
9
40±24
46
3
1
2.4±0.3
84±57
16±4.4
3.2±0.3
27
5
24
56
nd
-1
Eu (Bq kg )
n
241
5.3±1.0
-1
n
155
Upstream Process Outfall Footprint Downstream
(n=91)
(n=83)
(n=5)
(n=35)
Am (Bq kg-1)
n
nd
Figure 1
137
Cs in Ottawa River sediment.
nd
Figure 2
241
Am in Ottawa River sediment.
Am-241 is another anthropogenic radionuclide that was
globally distributed by atmospheric testing and was detected at most of the deeper sites. Upstream concentrations are
not significantly different than downstream concentrations
(see Table 1), with highest concentrations at the process
outfall, and rapidly decreasing in the footprint (see Figure
2). Concentrations of 241Am are about 40-fold lower than
those of 137Cs (see Table 1).
68
Co-60 is an activation product and present in Ottawa River
sediments due to operations at CRL. 60Co is only found at
upstream sites within 2 km of the process outfall, again suggesting limited upstream transport of CRL radionuclides
(see Table 1 and Figure 3). At the process outfall, concentrations of 60Co are similar to those of 137Cs, but 60Co concentrations decline much more rapidly through the footprint
(see Figure 3). Downstream concentrations of 60Co are
about 10-fold lower than 137Cs (see Table 1).
Figure 3
60
Co in Ottawa River sediment.
aecl Nuclear Review
vol 1, Number 1, june 2012
anthropogenic radionuclides in ottawa river sediment near
chalk river laboratories - d.j. rowan
3.2 Primordial Radionuclides
A number of primordial radionuclides were detected
in Ottawa River sediment, including K-40 and radionuclides from the 238U and Th decay series. K-40 is the
most abundant of the primordial radionuclides, averaging 537±5 Bq kg-1. K-40 concentrations in Ottawa River
sediment are not affected by CRL effluent, with very similar concentrations at all sites (see Figure 6). K-40 concentrations exceed those of all anthropogenic radionuclides except at some process outfall and footprint sites.
Figure 4
152,154,155
Eu in Ottawa River sediment.
Bi-214 (238U decay series) and 228Ac (232Th decay series) also
show no impact of CRL operations, with uniform concentrations at all sites (see Figure 7). These radionuclides occur
at much lower concentrations in Ottawa River sediment
than 40K, with 214Bi averaging 16.1±0.3 Bq kg-1 and 228Ac averaging 33.3±0.9 Bq kg-1. These primordial radionuclides
Figure 5
94
Nb in Ottawa River sediment.
Figure 6
40
K in Ottawa River sediment.
Eu, 154Eu and 155Eu were detected at some sites near the
process outfall, within the footprint and downstream. 152Eu
was the most widely detected of the three Eu radioisotopes,
with much fewer detections for 154Eu and 155Eu, which was
detected only in the immediate vicinity of the process outfall (see Table 1). Concentrations of Eu radioisotopes drop
off very rapidly from the process outfall through the footprint, and are present at very low concentrations at downstream sites (see Figure 4).
152
Nb was detected at low concentrations in the vicinity of
the process outfall, dropping off to very low downstream
concentrations of a few Bq kg-1 (see Table 1 and Figure 5).
As with other CRL derived radionuclides, there is evidence
of limited upstream transport (see Figure 5). The distribution of 94Nb is similar to 241Am and 152Eu, and 94Nb was detected primarily in depositional sediments at deeper sites.
94
Figure 7
214
Bi and 228Ac in Ottawa River sediment.
AECL NUCLEAR REVIEW
69
aecl Nuclear Review
vol 1, Number 1, june 2012
are present in concentrations similar to 60Co at downstream sites and are about 10-fold higher than the other anthropogenic radionuclides at downstream sites.
Conclusions
Cs-137 and 241Am in Ottawa River sediment are due to
weapon-test fallout and CRL releases. CRL operations have
about doubled weapon test fallout 137Cs at downstream
sites. Am-241 occurs at much lower concentrations and
appears to be retained near the process outfall, with no
significant difference between upstream and downstream
(>2 km from the process outfall) sites. Other anthropogenic radionuclides detected in Ottawa River sediment
include 60Co, 94Nb, 152Eu, 154Eu and 155Eu. These radionuclides are also found at some sites within 2 km upstream
of the process outfall, suggesting upstream transport of
small amounts of CRL releases. Upstream movement of
surface water during periods of strong upstream wind has
been previously observed [7]. These radionuclides decline rapidly with distance from the process outfall, and as
with 137Cs and 241Am, reach stable levels after about 2 km
References
anthropogenic radionuclides in ottawa river sediment near
chalk river laboratories - d.j. rowan
downstream of the process outfall. 137Cs and 94Nb decline
to a lesser degree from process outfall to downstream sites,
suggesting that these radionuclides are more mobile that
the other anthropogenic radionuclides.
Comparison of anthropogenic radionuclides with several
representative primordial radionuclides shows that with
the exception of sites at the process outfall and within 2 km
downstream of the process outfall, primordial radionuclide
concentrations greatly exceed CRL derived anthropogenic
radionuclide concentrations. In fact, CRL derived 137Cs in
sediment at downstream sites is only about 20% of naturally occurring 40K. Co-60 concentrations in sediment at
downstream sites are only about 3.5% of 40K and are similar
to representative 238U and 232Th decay series radionuclides.
Other anthropogenic radionuclides occur in much lower
concentrations (<1% of 40K) and are about an order of magnitude lower than representative 238U and 232Th decay series
radionuclides. Thus, over 60 years of radionuclide releases
from operations at CRL have had little impact on radionuclide concentrations in Ottawa River sediment, except
at a few sites immediately adjacent to the process outfall.
[1] D.J. Rowan and J.B. Rasmussen, 1994, “Bioaccumulation of Radiocesium by Fish: The Influence of Physicochemical Factors and Trophic Structure”. Canadian Journal of Fisheries and Aquatic Sciences, 51(11), pp. 2388-2410
[2] D.J. Rowan and J.B. Rasmussen, 1996, “Measuring the Bioenergetic Cost of Fish Activity in Situ Using a Globally Dispersed Radiotracer (137Cs)”. Canadian Journal of Fisheries and Aquatic Sciences,
53(4), pp. 734-745
[3] D.J. Rowan, 2012 In press. “Bioaccumulation Factors and the Steady State Assumption for Cesium Isotopes in Aquatic Foodwebs Near Nuclear Facilities”. Journal of Environmental Radioactivity.
[4] F.W. Whicker, C.T.Garten Jr., D.M. Hamby, K.A. Higley, T.G. Hinton, D.I. Kaplan, D.J. Rowan and R.G. Schreckhise, 2007, “Cesium-137 in the Environment: Radioecology and Approaches to Assessment
and Management”. National Council on Radiation Protection and Measurements, Bethesda, NCRP Report No. 154
[5] L.J. Jackson, D.J. Rowan, R.J. Cornett and J. Kalff, 1994, “Myriophyllum Spicatum Pumps Essential and Nonessential Trace Elements from Sediments to Epiphytes”. Canadian Journal of Fisheries
and Aquatic Sciences, 51(8), pp. 1769-1773
[6] F.H. Rigler, 1952. “Study of Radioactivity in Ottawa River Organisms and Bottom Deposits”. M.A. Thesis, University of Toronto
[7] W.F. Merritt, 1964, “Studies of Dilution in the Ottawa River Using Rhodamine B 2 - CRNL to Pembroke”. Atomic Energy of Canada Ltd. Report, AECL 2030
An official publication of
Atomic Energy of Canada Limited
www.aecl.ca
AECL NUCLEAR REVIEW
71
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement