Nuclear Engineering and Design 311 (2017) 199–212

Contents lists available at ScienceDirect

Nuclear Engineering and Design

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

Optimization of plutonium and minor actinide transmutation in an

AP1000 fuel assembly via a genetic search algorithm
J. Washington, J. King ⇑
Nuclear Science and Engineering Program, Colorado School of Mines, 1500 Illinois St., Golden, CO 80401, USA

h i g h l i g h t s g r a p h i c a l a b s t r a c t

We model a modified AP1000 fuel

assembly in SCALE6.1.

We couple the NEWT module of
SCALE to the MOGA module of
DAKOTA.

Transmutation is optimized based on
choice of coating and fuel.

Greatest transmutation achieved with
PuZrO2MgO fuel pins coated with
Lu2O3.

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

Article history: The average nuclear power plant produces twenty metric tons of used nuclear fuel per year, which con-
Received 21 May 2016 tains approximately 95 wt% uranium, 1 wt% plutonium, and 4 wt% fission products and transuranic ele-
Received in revised form 23 November 2016 ments. Fast reactors are the preferred option for the transmutation of plutonium and minor actinides;
Accepted 28 November 2016
however, an optimistic deployment time of at least 20 years indicates a need for a near-term solution.
Available online 18 December 2016
Previous simulation work demonstrated the potential to transmute transuranic elements in a modified
light water reactor fuel pin. This study optimizes a quarter-assembly containing target fuels coated with
spectral shift absorbers for the transmutation of plutonium and minor actinides in light water reactors.
The spectral shift absorber coating on the target fuel pin tunes the neutron energy spectrum experienced
by the target fuel. A coupled model developed using the NEWT module from SCALE 6.1 and a genetic
algorithm module from the DAKOTA optimization toolbox provided performance data for the burnup
of the target fuel pins in the present study. The optimization with the coupled NEWT/DAKOTA model pro-
ceeded in three stages. The first stage optimized a single-target fuel pin per quarter-assembly adjacent to
the central instrumentation channel. The second stage evaluated a variety of quarter-assemblies with
multiple target fuel pins from the first stage and the third stage re-optimized the pins in the optimal sec-
ond stage quarter-assembly. An 8 wt% PuZrO2MgO inert matrix fuel pin with a 1.44 mm radius and a
0.06 mm Lu2O3 coating in a five target fuel pin per quarter-assembly configuration represents the optimal
combination for the transmutation of plutonium and minor actinides in the LWR environment considered
in this study.
Ó 2016 Elsevier B.V. All rights reserved.

1. Introduction

An average Light Water Reactor (LWR) nuclear power plant pro-

duces 20 metric tons of used nuclear fuel per year (Department of
⇑ Corresponding author.
Energy, 2002). This used fuel consists of approximately 95 wt% ura-
E-mail addresses: [email protected] (J. Washington), [email protected]
(J. King). nium, 1 wt% plutonium, and 4 wt% fission products and other

http://dx.doi.org/10.1016/j.nucengdes.2016.11.030
0029-5493/Ó 2016 Elsevier B.V. All rights reserved.
200 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212

transuranic elements (Department of Energy, 2002). The majority 2.1. Transmutation fuels
of the residual heat produced by used fuel in the first 100 years
of storage originates from the radioactive decay of the fission prod- The use of plutonium fuel in light water reactors may be consid-
ucts. After 100 years, the transuranic elements are responsible for ered as a partial solution for used nuclear fuel disposition and the
the majority of the heat production. This decay heat must be con- use of plutonium as a fuel in a nuclear reactor is an alternative to
sidered in the design of both short and long-term storage facilities vitrification and disposal (Galperin, 1995). Currently, only Mixed
(Wigeland et al., 2006). The design basis lifetime of such facilities Oxide fuel (MOX) makes use of plutonium as a component of trans-
could be greatly reduced if the actinide components of used fuel mutation fuel in LWRs (Galperin, 1995). Plutonium may be
did not require long-term storage or disposal (Wigeland et al., retrieved from used nuclear fuel through a standard PUREX repro-
2006). The separation of plutonium from used fuel reduces the cessing method which can separate uranium and plutonium from
repository heat load and provides an additional source of fissile used fuel and provides the source material for recycled or transmu-
and fertile materials to the nuclear fuel cycle. tation fuel (Özdemir et al., 2011). Though MOX fuel has been
Fast reactors are generally preferred for the destruction of plu- approved for use in several PWR reactors in the United States, it
tonium and minor actinides; however, an optimistic deployment is currently only routinely used in reactors in other countries such
time of at least 20 years for these reactors creates a need for as France and Japan (Trellue, 2006).
nearer-term solutions (Salvatores and Palmiotti, 2011). As an A previous study evaluated sixteen different target fuels, both
interim solution, various studies have assessed the feasibility of with and without initial minor actinides, in a Westinghouse
plutonium and minor actinide transmutation in existing LWRs in AP1000 XL Robust Fuel Assembly (Washington et al., 2017). The
order to reduce the stockpile of these materials (Salvatores and sixteen target fuels were evaluated over a 1400 Effective Full
Palmiotti, 2011; Washington et al., 2017). Previous work has incor- Power Day (EFPD) depletion period to contrast their relative rates
porated plutonium, both with and without minor actinides, into of plutonium transmutation, curium production, and total minor
reactors as a target fuel pin (see Section 2.1). Other options include actinide transmutation. Though some of these fuels had been pre-
full assemblies of target fuel pins or whole reactors fueled with viously evaluated for their transmutation performance, no single
innovative fuel types, such as plutonium zirconium hydride with study had evaluated more than three different transmutation fuels.
thorium hydride or plutonium hydrides with uranium hydrides These transmutation fuels all contained plutonium, though some
in a zirconium matrix. studies included an initial quantity of minor actinides. The target
The objective of this study is to optimize a fuel assembly for the fuels were evaluated both with and without minor actinides to
purposes of plutonium and minor actinide transmutation in exist- determine how the addition of these materials changed the overall
ing LWRs. Maximizing the destruction of the fissile plutonium iso- plutonium transmutation properties of the fuels. The transmuta-
topes is as important as minimizing the end-of-life concentrations tion fuels were then considered with coatings of various spectral
of the minor actinides. A spectral shift absorber applied to the shift absorbers to determine which coating provided the greatest
outer diameter of plutonium fuel can improve the plutonium, nep- improvement in overall transmutation performance.
tunium, americium, and curium transmutation performance. A The three fuels evaluated in the present study, 5 wt% MOX
spectral shift absorber is a material that absorbs neutrons of a (MOX (5% PuO2)), plutonium 60 wt% thorium zirconium hydride
specific energy window thereby adjusting the overall neutron (Pu0.31ZrH1.6Th1.08), and 8 wt% plutonium magnesium oxide inert
energy spectrum seen by the material shielded with the spectral matrix (PuZrO2MgO), are selected from the previous study
shift absorber, in this case a target fuel. Spectral shift absorbers (Washington et al., 2017). The inert matrix fuels (Pu0.31ZrH1.6Th1.08
may be used to shift the effective neutron energy spectrum of a and PuZrO2MgO) are fueled with plutonium and minor actinides
reactor core by increasing the absorption of neutrons in a certain (americium, neptunium, and curium). The 5 wt% MOX fuel was
energy range, at the penalty of decreasing the overall flux selected for its transmutation performance and its commonality
(Duderstadt and Hamilton, 1976). with fuels currently in use. The 60 wt% thorium hydride fuel
Determining the optimal fuel, spectral shift absorber coating, achieved the greatest quantity of transmutation of the potential
and assembly geometry for a LWR transmutation scheme is a thorium hydride fuels and also broadens the variety of fuels eval-
challenging design problem. A genetic search algorithm can eval- uated. The 8 wt% PuZrO2MgO fuel resulted in the greatest quantity
uate a large variety of possible design parameters to determine a of plutonium and minor actinide transmutation of the fuels evalu-
solution that maximizes plutonium transmutation and minimizes ated in the previous study.
minor actinide production. A previous study recommended a set Previous studies evaluated eleven different burnable absorbers
of transmutation target fuels and spectral shift absorbers for as spectral shift absorber coatings (Washington et al., 2014,
use in a modified AP1000 fuel assembly (Washington et al., 2017). Previous work (Washington et al., 2014) incorporated the
2017). As a learning program, the selected genetic algorithm is eleven different potential spectral shift absorbers as a layer in
able to determine global and local solutions to the design modified TRISO fuel particles to act as burnable absorbers to
problem that may have been overlooked by previous manual reduce initial excess reactivity in a very high temperature reactor.
optimizations. The requirements for melting temperature, natural abundance of
absorbing isotope, and extended decay time still apply to this
study even though the initial evaluation occurred in a different
reactor environment. The present study considers four of these
2. Background spectral shift absorbing materials, Gd2O3, In2O3, Lu2O3, and
Sm2O3 (Table 2.1). A previous study (Washington et al., 2017)
This section provides background information on the use of achieved the highest rate of plutonium transmutation with the
transmutation fuels and the use of genetic search algorithms for lowest production of curium-244 using these four spectral shift
nuclear reactor core design. A previous paper selected the trans- absorbers.
mutation fuels for their overall performance in plutonium and
minor actinide transmutation (Washington et al., 2017). These 2.2. Optimization and genetic algorithms
fuels are coated with spectral shift absorbers to increase the rela-
tive fast flux in the target fuels to mimic a fast reactor Design problems, especially models of complex systems, gener-
environment. ally consider more than a single measure of performance (Eddy
 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212 201

and Lewis, 2001). These objectives may compete, such that an and replacement portion of the algorithm (Eddy and Lewis,
improvement in one objective comes at the expense of another 2001). The population is then sorted into a data set (mating pool)
(Eddy and Lewis, 2001). In a problem with two or more objectives, where solutions with a higher objective function evaluation are
at least two of which compete, there is typically no single optimal represented more times than the solutions with a lower objective
design but rather, a range of designs which satisfy the design crite- function evaluation (Adams et al., 2013; Eddy and Lewis, 2001).
ria (Baraldi et al., 2009; Eddy and Lewis, 2001). These non- In order to ensure diversity of the solutions, the mating pool does
dominated solutions are referred to as a Pareto set and the range not exclude solutions and each feasible solution is given an equal
of options in the set are called Pareto optimal solutions (Eddy chance of reproduction (Eddy and Lewis, 2001). The individual
and Lewis, 2001). The Pareto set consists of all the solutions to a solutions are chosen at random and matched based on the user
given problem that satisfy the optimization conditions. Solutions determined population size, the number of children per crossover,
to the optimization conditions which cannot be improved upon, and the crossover rate (Adams et al., 2013; Eddy and Lewis, 2001).
lying along the Pareto frontier, are considered Pareto optimal solu- Three genetic operators (crossover, mutation and selection) are
tions. The Pareto set may contain a significant number of Pareto the primary exploration mechanisms in the search algorithm (Gen
optimal solutions, along with a number of dominant solutions et al., 2008; Pereira and Sacco, 2008). Crossover may take place in a
which must be evaluated and sorted (Eddy and Lewis, 2001). One number of ways and generally swaps one solution parameter with
efficient way to generate a fully populated Pareto set is the use another from a different parent solution to produce the offspring
of a Multi-Objective function Genetic search Algorithm (MOGA) solutions (Adams et al., 2013; Aghaie et al., 2013). The solutions
(Baraldi et al., 2009; Eddy and Lewis, 2001). A genetic algorithm which are selected for crossover are called a mating pair and pro-
determines an optimal solution set by emulating the process living duce offspring or children (Eddy and Lewis, 2001). Crossover emu-
creatures use to evolve (Aghaie et al., 2013; Gen et al., 2008). lates the exchanged of genetic material between parents to
Genetic algorithms do not need to consider the gradient informa- produce offspring.
tion which many other optimization methods require (Eddy and The children from the crossover are then subjected to a muta-
Lewis, 2001; Jiejuan et al., 2004). This capability makes it possible tion step (Eddy and Lewis, 2001; Pereira and Sacco, 2008). The
for the genetic search algorithm to avoid being trapped in locally mutation operator produces random or semi-random (a random
optimal solutions by high gradient functions (Eddy and Lewis, event that can be recreated given an initial seed) changes in the
2001; Jiejuan et al., 2004). existing population (Baraldi et al., 2009; Gen et al., 2008). The
A genetic algorithm usually follows a consistent process regard- mutation encourages the exploration of the problem space, intro-
less of the design or structure of the program (Eddy and Lewis, duces new genes which were not present in the initial population,
2001; Pereira and Sacco, 2008). The process has five basic compo- and re-introduces genes which were previously lost by the popula-
nents: a representation of the problem, a way to create an initial tion for evaluation in a new configuration (Gen et al., 2008;
population, the ability to evaluate the fitness of potential solutions, Norouzi et al., 2011; Pereira and Sacco, 2008).
a method to alter the offspring, and the parameters which govern Selection is the final stage of the exploration step of the genetic
the genetic algorithm (Gen et al., 2008; Pereira and Sacco, 2008). search algorithm (Gen et al., 2008; Pereira and Lapa, 2003). After
The first step of a genetic search algorithm is to determine, or be the offspring have been produced by the crossover and mutation
given, an initial population with design variable values between operators they are evaluated against the fitness function and sub-
the design variable boundaries (Eddy and Lewis, 2001; Pereira jected to a selection operator to determine the offspring which are
and Lapa, 2003). These initial designs, if they are randomly gener- allowed to reproduce (Gen et al., 2008; Pereira and Lapa, 2003).
ated, may be tested against the other initial designs to eliminate The pressure applied by the selection operator is generally low
any duplicates among the initial population to ensure a greater during the beginning of the genetic search, to encourage explo-
diversity of potential solutions (Adams et al., 2013; Eddy and ration, and increases as the search progresses to narrow in on an
Lewis, 2001). This clone detection is useful for several reasons: it optimal solution (Gen et al., 2008; Pereira and Sacco, 2008).
avoids multiple evaluations of the same solution, it encourages
population of the solution space, the lack of duplicate solutions 2.3. Genetic algorithms applied to nuclear systems
saves on storage and, lastly, it helps to avoid incorrect convergence
on a locally optimal solution (Adams et al., 2013; Eddy and Lewis, Genetic search algorithms have been used to devise fuel loading
2001). patterns for nuclear power reactors, as the usual computational
Once a population is determined, each solution is tested against methods for determining a core loading pattern may take weeks
a fitness function (Aghaie et al., 2013; Eddy and Lewis, 2001). The to find a satisfactory arrangement of new and used fuel assemblies
fitness of a population member is determined by its objective func- (Zhao et al., 1998). The potential number of loading patterns for a
tion (or functions) (Aghaie et al., 2013; Gen et al., 2008). The objec- PWR, considering one-eighth symmetry, is in the range of 1026
tive functions evaluate the fitness of each individual, but may vary (Alim et al., 2008). Fuel loading parameters include the arrange-
in range over different problem spaces and are usually normalized ment of fresh fuel, the arrangement of the used fuel, and then
from zero to one for consistency (Gen et al., 2008). Fitness func- the placement of burnable absorbers into fresh fuel locations to
tions contain constraints which are computed for each lower the power peaking in regions where it exceeds the limiting
solution and penalize infeasible solutions during the selection values (Yilmaz et al., 2006a). The multi-objective potential of

Table 2.1
Spectral shift absorbers considered in the present study.

Absorber Density (g/cm3) @ 1200 K Melting point (K) Isotope of interest Absorption cross-section ra (barns) Resonance integral (barns)
Gd2O3 7.35 2612a 157
Gd 254,000b 700b
In2O3 7.12 2185a 115
In 215b 3300b
Lu2O3 9.34 2700a 176
Lu 2090b 1087b
Sm2O3 7.54 2542a 149
Sm 40,140b 3390b
a
Lide (2009).
b
Dunford and Kinsey (1998).
202 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212

genetic algorithms is well-suited for this problem (Zhao et al., plutonium and minor actinide transmutation, as well as the bur-
1998). Objectives used in the determination of a Pareto set for fuel nup lifetime of the assembly. The script then feeds the target fuel
loading include cycle length, neutron leakage, burnup, and power pin designs back to the neutronics code for further evaluation.
peaking (Norouzi et al., 2011; Zhao et al., 1998). Since reactor cores A model developed using the NEWT module of the SCALE 6.1
are generally symmetric, various slices, from one-half to one- code using the ENDF/B-VII.0 cross-section libraries provided per-
eighth of a reactor core, may be used in the search algorithm to formance data for the burnup of the target fuel pins in the present
decrease computational time (Norouzi et al., 2011; Zhao et al., study (Chadwick et al., 2006; Oak Ridge National Laboratory,
1998). One test case, with one-sixth symmetry, produced Pareto 2011). The NEWT model is a two-dimensional, discrete-ordinate,
optimal solutions within 20 generations with a computational time reflected model using 16th order symmetric quadrature. The
of two hours; however, the problem was only optimized based on model used a coarse mesh of 34 by 34 cells for the overall
power peaking with the given assemblies (Norouzi et al., 2011). quarter-assembly, with target fuel pin meshes of 20 by 20 for each
The use of a genetic search algorithm may also be applied to the target fuel pin. Each cell is further overlaid with a four by four
placement of burnable absorbers in a nuclear reactor core (Alim mesh, which may be condensed by SCALE if adjacent cells are
et al., 2008). In one study, the number and concentration of the indistinguishable from each other. The driver fuel pins each con-
burnable absorber pins in a reactor core were determined by a tain 3 weight percent (wt%) enriched uranium dioxide fuel
genetic algorithm, which could be done in conjunction with the (Schulz, 2006; Westinghouse Electric Corp et al., 1992). The NEWT
determination of the fuel loading pattern or conducted as a sepa- sequence employed the prepackaged 238 group ENDF/B-VII.0
rate search (Alim et al., 2008). Core loading patterns tend to be lim- cross-section library with reflective quarter-assembly boundary
ited by a smaller number of potential burnable absorber loading conditions. The burnup model is simulated over 1400 EFPD, assum-
options which have been predetermined by the reactor design ing a total assembly power of 21.66 MWth based on an even distri-
(Yilmaz et al., 2006a). The genetic search expanded the potential bution of 3400 MWth across the 157 fuel assemblies of an AP1000
placement schemes and therefore the allowable fuel loading pat- (Schulz, 2006). The target fuels, uranium fuel pins, and the spectral
terns (Yilmaz et al., 2006a). The use of genetic algorithms may be shift absorbers are depleted during the burnup simulation.
treated as a black box approach by the user when additional crite- Table 3.1 lists the dimensions of the fuel assembly and compo-
ria are applied to the fitness functions to generalize the search nents included in the model shown in Fig. 3.1. The radius of the
options (Yilmaz et al., 2006a). fuel pellets was 0.410 cm with a gap thickness of 0.017 cm, a ZIR-
Yilmaz et al. (2006b) evaluated a burnable absorber scheme for LOTM cladding thickness of 0.057 cm, and a fuel pin pitch of
pressurized water reactors with fuel assemblies containing three 1.260 cm (Schulz, 2006). A standard 17  17 Extra-Large (XL)
different gadolinia (Gd2O3) enrichments (100 wt% enriched Robust Fuel Assembly is 21.417 cm on a side and contains 264 fuel
gadolinium-155, 100 wt% enriched gadolinium-157, and 70 wt% pins with 24 control pin channels and one central instrumentation
enriched gadolinium-157 with 30 wt% gadolinium-155). The channel (Schulz, 2006). The calculations in this study consider one
15  15 fuel assembly in this evaluation contained a mixture of quarter of an assembly to reduce the computational time required
UO2 fuel and Gd2O3 absorber pins. A genetic algorithm evaluated for convergence.
various placement schemes to optimize the quantity of burnable A fraction of the plutonium in the minor actinide bearing fuels
absorber in the reactor (Yilmaz et al., 2006b). The genetic algo- is replaced by a mixture of minor actinides (americium, curium,
rithm produced a geometry which resulted in a reduction of the and neptunium), based on isotope ratios determined for an average
required gadolinium mass from 31.1 kg to 2.7 kg in the 100 wt% PWR fuel assembly with a burnup of 41,200 MWd/MTHM at
enriched gadolinium-155 case (Yilmaz et al., 2006b). 3.75 wt% enrichment after 23 years of decay (Department of
Generally, previous studies on the optimization of nuclear reac- Energy, 2002). Table 3.2 shows the composition of the minor acti-
tor cores developed individually tailored genetic algorithms to nide containing target fuels used in this study and Table 3.3 shows
determine the Pareto optimal solutions (Alim et al., 2008; the initial isotope concentrations assumed in the reprocessed plu-
Norouzi et al., 2011; Yilmaz et al., 2006a). Each tailored program tonium used in the target fuels.
contained the standard genetic algorithm components, such as The four spectral shift absorbers used in this study are shown in
crossover and mutation, arranged with various start and stop con- Table 2.1 along with their densities, melting points, and con-
ditions to quickly terminate infeasible designs in order to reduce stituent isotopes with the highest cross-section or resonance inte-
computational time (Alim et al., 2008; Norouzi et al., 2011; gral. The genetic search algorithm varied the position and number
Yilmaz et al., 2006a). In the present study, a coupled nuclear reac- of target fuel pins as well as the internal geometry of the target fuel
tor simulation code and a genetic search algorithm optimized the pins, including fuel pellet diameter and absorber coating thickness.
reactor assembly geometry for the transmutation of plutonium
and minor actinides. This automated method explored the signifi-
cant problem space represented in the number of spectral shift Table 3.1
absorbers, target fuel types, possible assembly configurations, AP1000 Robust fuel assembly parameters used in the present study (Schulz, 2006).
and potential target fuel pin geometries.
Component Parameter Composition (wt%) Density
Fuel pellet 0.410 cm 3 wt% enriched 10.96 g/cm3
radius uranium oxide
3. Computational models
Gap thickness 0.017 cm Vacuum 0 g/cm3
radius
For this study, a coupled neutronics model (NEWT) (Oak Ridge Clad thickness 0.057 cm 1.25 wt% Nb, 1.1 wt% Sn, 6.56 g/cm3
National Laboratory, 2011) and genetic search algorithm (DAKOTA) radius 0.105 wt% Fe,
0.055 wt% Ni
(Adams et al., 2013) determined configurations of target fuels,
0.055 wt% Cr, 0.00022
spectral shift absorbers, and positions of the target fuel pins for wt% C, 97.43478 wt% Zr
optimal transmutation performance in an AP1000 Robust Fuel Fuel pin pitch 1.260 cm
assembly. The two codes are coupled through a PERL script that Assembly pitch 21.417 cm
extracts the burnup data from the neutronics model and feeds this Assembly 21.66 MW
thermal
information to the genetic algorithm. The algorithm takes the bur-
power
nup results and contrasts each burnup run to determine the total
 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212 203

Fig. 3.1. NEWT quarter-assembly model.

The gap and cladding thickness of the modified AP1000 fuel pins Table 3.3
were held constant, although the outer diameter of the modified Reprocessed plutonium composition used in this study.
fuel pins was allowed to change based on the dimensions of the Isotope Weight percent
internal components. To reduce the computational overhead, the Plutonium-238 2.37
simulations were performed in three stages (described in Section 4) Plutonium-239 61.38
using the Multi-Objective Genetic Algorithm (MOGA) module from Plutonium-240 25.97
DAKOTA (Adams et al., 2013). The algorithm produced 600 off- Plutonium-241 4.81
Plutonium-242 5.47
spring from an initial population of 50 for 12 generations per opti-
mization stage. The NEWT model evaluated each parent and
offspring using the previously listed model parameters. Setting
the mutation and crossover rates of the MOGA package to 95% of
the first stage, the second stage varied the number of target fuel
their maximum value encouraged rapid diversification in the pop-
pins and their location within the quarter-assembly to achieve an
ulation and ensured a complete evaluation of the problem space.
optimum target arrangement. Based on this arrangement, the final
The DAKOTA optimizations included the three objective func-
stage re-optimized the geometry of the target fuel pins.
tions shown in Table 4.1. The first objective function, the EFPD pen-
alty, considers how long the quarter-assembly remained critical.
The second objective function compares the initial and final quan- 4.1. First stage optimization
tities of fissile plutonium. The final objective function considers the
initial and final quantities of minor actinides. The final two objec- The first stage optimization considered a single pin located
tive functions are normalized to the initial mass of the transmuta- adjacent to the central instrumentation channel in each quarter-
tion fuel in the target fuel pins to ensure that the comparison assembly. Scoping evaluations using the combined DAKOTA and
between two quarter-assemblies with different quantities of target NEWT codes, considering multiple fuel and spectral shift absorber
fuel was meaningful. The optimal values of all three objective func- configurations, indicated that this position (see Fig. 4.1) has a high
tions are zero. potential for successful transmutation. This position is also adja-
cent to the central instrumentation channel, close to where the
previous study evaluated the target fuels and spectral shift absor-
4. Optimization results bers (Washington et al., 2017). The selected position provided ini-
tial results similar to those of the previous study, confirming that
Performing the optimization in three stages further reduced the the combined model yields reproducible results.
computational time. The first stage adjusted the geometry of a sin- The diameter of the target fuel pellet and the thickness of the
gle target fuel pin in the quarter-assembly. Based on the pin from spectral shift absorber coating varied greatly throughout the simu-

Table 3.2
Target fuel composition with minor actinides at beginning of lifetime.

Constituents (wt%)
Fueld 232
Th Uc Pu 241
Am 243
Am 244
Cm 237
Np O Zr H Y/Mg
MOX (5% PuO2)a,c 0.00 83.72 3.73 0.37 0.06 0.01 0.27 11.86 0.00 0.00 0.00
Pu0.31ZrH1.6Th1.08b 59.97 0.00 14.98 1.48 0.22 0.04 1.06 0.00 21.83 0.39 0.00
PuZrO2MgO (8%) 0.00 0.00 6.73 0.66 0.10 0.02 0.48 28.93 40.87 0.00 22.19

All O, Mg, Zr, H, and Th elements are in their naturally occurring state.
a
MOX properties are from Trellue (2006).
b
Zirconium hydride compositions are from Olander et al. (2009).
c
Uranium is natural uranium without enrichment or depletion.
d
Plutonium concentrations shown in Table 3.3.
204 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212

Table 4.1 Table 4.2 lists the five near optimal designs produced by the
Objective functions used in the present genetic algorithm. combined SCALE and DAKOTA setup in the first optimization stage.
Objective function The top five designs in this stage all have the same spectral shift
1 1400-Effective full power days absorber and target fuel type. Table 4.3 shows the distribution of
2 (Pu-239f + Pu-241f)/(Pu-239i + Pu-241i) the target fuels and spectral shift absorbers during this optimiza-
3 (Np-237f + Cm-242f + Cm-243f + Cm-244f + Am-241f + Am-243f)/(Np- tion stage. The lutetium oxide coating around the inert matrix fuel
237i + Cm-242i + Cm-243i + Cm-244i + Am-241i + Am-243i) is the most commonly selected combination. The near optimal
Isotopei = Mass of isotope at beginning of life. designs resulted in a minimum plutonium decrease of 88 wt%
Isotopef = Mass of isotope at end of life. and reduced the initial minor actinide content of the target fuel
by 48 wt% or more. Each of these simulated fuel quarter-
lations, in order to fully explore the problem space. Fig. 4.1 displays assemblies achieved a full 1400 EFPD lifetime, minimizing the
an example of one of the ways the coupled NEWT/DAKOTA code operational lifetime penalty from the parasitic neutron absorbers
varied the target fuel pin geometry. In many cases throughout in the target fuel pin.
the instances of the optimization model, the spectral shift absorber
coating was too thick, reducing the overall lifetime of the quarter- 4.2. Second stage optimization
assembly by the absorption of a significant quantity of the neu-
trons in the reactor. These cases are rated as less competitive by The second optimization stage increased the quantity of target
the first objective function. The cases with a thick target fuel pellet fuel pins and allowed for different target positions within a
and thin spectral shift absorber coating provided a higher quantity quarter-assembly. This stage tested if variations in the position of
of initial reactivity, but were generally rated lower by the second the target fuel pins improved the transmutation performance of
objective function as they ended the quarter-assembly life cycle the quarter-assembly. This stage of the optimization fixed the tar-
with a higher quantity of plutonium than the thinner target fuel get fuel pin geometry, based on the first stage pin with the lowest
pellets. Thinner target fuel pellets work well due to the relative score for Objective Function 3 (Case 5 in Table 4.2), and optimized
reduction in self-shielding, allowing a higher percentage of minor the number and positions of these pins within the quarter-
actinide transmutation than thicker target fuel pellets with a assembly. The majority of the possible configurations in this stage
higher rate of self-shielding. Thin geometries result in both less result in quarter-assemblies that do not produce a critical reactor
plutonium at end of life as well as better minor actinide transmu- core. The following discussions omit these cases.
tation rates. Consequently, the genetic algorithm’s optimization Two metrics determined the optimal configuration of target fuel
functions rate these configurations as more effective. pins using the three objective functions. The first metric is the

Fig. 4.1. Optimization of one pin in a quarter-assembly configuration.

 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212 205

Table 4.2
The five near optimal single target fuel pin designs produced by the first stage optimization with the coupled DAKOTA/NEWT code.

Case Fuel pin radius (mm) Absorber thickness (mm) Spectral shift absorber Fuel type Objective function 2 Objective function 3
1 1
1 1.20  10 3.71  10 Lu2O3 PuZrO2MgO (8%) 0.082513 0.522594
2 8.83  102 8.42  101 Lu2O3 PuZrO2MgO (8%) 0.091562 0.501212
3 7.63  102 1.30 Lu2O3 PuZrO2MgO (8%) 0.105265 0.493356
4 3.63  101 8.42  101 Lu2O3 PuZrO2MgO (8%) 0.10584 0.490521
5 6.01  101 8.54  101 Lu2O3 PuZrO2MgO (8%) 0.117524 0.488797

EFPD penalty (Objective function 1 in Table 4.1), and the second additional target fuel pin added to the quarter-assembly configura-
metric is the ratio of plutonium to curium transmutation (the ratio tion reduces the total reactivity available to the quarter-assembly.
of Objective functions 2 and 3 in Table 4.1). The EFPD penalty The target fuel pins are also coated with a spectral shift absorber,
(Fig. 4.2a) is a measure of how long the quarter-assembly remained which acts as an additional layer of self-shielding. As the number
critical compared to an unmodified quarter-assembly, which lasts of self-shielded pins increases in the quarter-assembly, both the
1400 EFPD. SCALE6.1 reports reactivity results at the midpoint of quarter-assembly as a whole and the individual fuel pins become
the burnup step and requires discretized burnup steps resulting more opaque to neutrons. In this study, fuel quarter-assemblies
in the non-continuous distribution of EFPD penalties represented with more than twenty-six target fuel pins cannot remain critical
in Fig. 4.2. This difference is important because it indicates how for more than 100 days.
much reactivity the quarter-assembly would remove from a whole Fig. 4.2b and Table 4.4 contains two examples of quarter-
reactor. The EFPD penalty is also important in ensuring the assembly configurations where the same number of target fuel
quarter-assembly reaches a reasonable burnup before removal pins, or fewer, achieved a lower EPFD penalty based on their posi-
from the reactor. The normalized plutonium to curium ratios tion in the quarter-assembly. The two data points corresponding to
allowed a fair comparison between quarter-assemblies that con- three target fuel pins per quarter-assembly indicate that the posi-
tained different initial masses of these isotopes. tion of the target fuel pins is important. These two quarter-
Fig. 4.2a shows the EPFD penalty as a function of the number of assemblies contain the same number of similar target fuel pins
target fuel pins for the second stage quarter-assemblies that in different locations, but show a difference of at least a year in
achieved at least 100 days of operation. Fig. 4.2b displays the same terms of active lifetime. The second two data points of importance
information quarter-assembly configurations with 18 or fewer in Fig. 4.2b and Table 4.4 are the points at 15 and 16 target fuel
target fuel pins in Fig. 4.2a with the lowest EFPD penalty. The pins per quarter-assembly. Those examples reveal a quarter-
assemblies containing more than 18 target fuel pins per quarter- assembly configuration where more target fuel pins can achieve
assembly had EPFD penalties of at least 800 days and were omitted a smaller EFPD penalty. These two quarter-assembly configura-
from further consideration. None of the options had an EFPD pen- tions indicate that a quarter-assembly with more than three target
alty equal to zero but a small number of the potential configura- fuel pins may achieve the smallest EFPD penalty if the target fuel
tions in Fig. 4.2b have an EFPD penalty of less than 100 days. All pins are placed in the correct configuration.
of the configurations with an EFPD penalty of 100 days or less Fig. 4.3a and b shows the ratio of plutonium-to-curium trans-
had fewer than five target fuel pins in the quarter-assembly. The mutation as a function of the EFPD penalty and the number of tar-
target fuel pin configurations corresponding to the points shown get fuel pins in the quarter-assembly, respectively. Fig. 4.3a shows
in Fig. 4.2 are the closest to the goal for the end of life reactivity. the differences between various quarter-assembly configurations
Table 4.4 displays data for selected cases in the second optimiza- in terms of the ratio of plutonium to curium as a function of the
tion stage with an EPFD penalty of less than 800 days. There were EFPD penalty. A number of assemblies have the same EFPD penalty
multiple cases satisfying this criteria with six or fewer pins per with a 10% variation in the transmutation ratio. While the range of
quarter-assembly. The pairs of one through six pin cases in variation in the transmutation metrics of the system is small, this
Table 4.4 represent the cases with the highest and lowest Objective figure does show that improvements are possible with variation in
Function Two over Objective Function Three ratios for a given both the number and positions of the target fuel pins.
number of target pins pre quarter-assembly. The quarter- Fig. 4.3b and Table 4.4 illustrate the differences in the pluto-
assemblies in Fig. 4.2b that have an EFPD penalty less than nium/minor actinide ratio (Objective function 2/Objective function
465 days represent the configurations that are likely to achieve a 3 in Table 4.4) as a function of pin position with the same number
full burnup cycle with additional optimization, or that will have of target fuel pins. The variation in plutonium to minor actinide
the smallest impact on a reactor containing only a few target ratio occurs most frequently in the quarter-assemblies with six
quarter-assemblies. It is likely that these quarter-assemblies may or fewer target fuel pins. The differences in the quarter-assembly
be further improved, as their end of life reactivity was close to that configurations reveal that assemblies with up to sixteen target fuel
of the full burnup of a standard AP1000 fuel quarter-assembly, or pins can achieve similar transmutation metrics as those assemblies
similar numbers of target fuel pins achieved a lower EFPD penalty with three target fuel pins. Of particular note is the five target fuel
based on their relative position in the quarter-assembly. pin quarter-assembly, which achieves a transmutation perfor-
Since the target fuel pins contain less fissile material than the mance metric similar to those of the one and two target fuel pin
uranium fuel pins used in the AP1000 fuel quarter-assembly, each quarter-assemblies.

Table 4.3
Frequency of each fuel type and spectral shift absorber during the single-pin DAKOTA evaluations.

Fuel type Count Frequency SSA type Count Frequency

PuZrO2MgO (8%) 290 45.31% Gd2O3 128 20.00%
Pu0.31ZrH1.6Th1.08 248 38.75% In2O3 152 23.75%
MOX (5% PuO2) 102 15.94% Lu2O3 178 27.81%
Sm2O3 182 28.44%
206 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212

a) EFPD penalty as a function of number of target fuel pins

b) EFPD penalty as a function of number of target fuel pins for

selected cases from the stage two evaluations
Fig. 4.2. EFPD penalty as a function of the number of target fuel pins per quarter-assembly.

Table 4.4
Selected stage two cases with an EPFD less than or equal to 725 days.

Number of target fuel pins Objective function 1 Objective function 2 Objective function 3 (Objective function 2)/
per quarter-assembly (Objective function 3) ratio
1 75 0.1518 0.5311 0.2858b
1 75 0.1038 0.4909 0.2115c
2 75 0.1425 0.5230 0.2725b
2 75 0.1124 0.4985 0.2254c
3 425 0.2362 0.5934 0.3980b
3 75 0.1268 0.5105 0.2484c
4 425 0.2325 0.5913 0.3933b
4 425 0.1482 0.5251 0.2822c
5 425 0.2320 0.5911 0.3925b
5a 425a 0.1647a 0.5398a 0.3051a,c
6 425 0.2125 0.5773 0.3681b
6 425 0.2035 0.5706 0.3566c
7 425 0.2272 0.5880 0.3863
8 425 0.2295 0.5901 0.3889
9 425 0.2164 0.5807 0.3727
11 425 0.2326 0.5933 0.3921
12 425 0.2298 0.5918 0.3883
13 425 0.2344 0.5955 0.3936
15 775 0.3204 0.6789 0.4719
16 425 0.2315 0.5939 0.3898
18 775 0.3679 0.7059 0.5213

Bold values show the case selected for the following optimization stage.
a
Selected for final optimization stage.
b
Highest objective 2/3 ratio with the same number of pins per quarter-assembly.
c
Lowest objective 2/3 ratio with the same number of pins per quarter-assembly.
 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212 207

The quarter-assemblies containing five target fuel pins shown target fuel pins have the same geometry in any given quarter-
in Fig. 4.3 and Table 4.4 have EFPD penalties similar to the assembly and varied greatly in dimension over the course of the
worst-case quarter-assemblies containing one or two target fuel simulations (see Fig. 4.5).
pins. Irradiation of the worst-performing of the quarter- The optimal transmutation would occur in a quarter-assembly
assemblies containing five target fuel pins achieved minor actinide configuration that produced the highest rates of plutonium and
transmutation metrics similar to those of the worst-performing of overall minor actinide transmutation with the lowest production
the single target pin quarter-assembly, 0.54 compared to 0.53, of curium and the lowest EFPD penalty. The best transmutation
respectively. The best-performing quarter-assembly containing results in this stage come from thin target fuel pins with a thin
five target fuel pins transmuted 17 wt% less actinides than the spectral shift absorber coating. The series of optimization runs in
best-performing quarter-assembly containing six target fuel pins. the final stage resulted in target fuel pins that were thicker than
The best-performing quarter-assembly with five target fuel pins the optimized single pin from Section 4.1 by an average factor of
reaches a lifetime of 975 EFPD, and may be able to be further opti- 3.15 (1.44 cm for Case 8 in the final optimization stage (Table 4.5)
mized. Based on the low plutonium to minor actinide ratios, the compared to 0.12 cm for Case 1 in the first optimization stage
similarity to the best performing single target pin fuel quarter- (Table 4.2)).
assembly, and the high end of cycle reactivity, this study selected Table 4.5 presents the nine near optimal configurations deter-
the quarter-assembly configuration in Fig. 4.4 for the target pin mined by the final optimization stage. At this stage in the opti-
reconsideration in the final optimization stage. mization all of the near optimal target fuel pin designs achieved
the minimum EFPD penalty and for this reason the results from
4.3. Third stage optimization Objective function 1 are not reported. In contrast to the first stage
optimization, that selected lutetium oxide in the top five cases, the
The third and final optimization stage used the optimized spectral shift absorber choices in the top nine cases in the final
quarter-assembly geometry from the second stage (Fig. 4.4) and stage included lutetium, samarium, and gadolinium oxides. Similar
reconsidered the target fuel pin configuration. This stage set the to the first stage of the optimization, the final stage selected the
location of five identical target fuel pins based on the results of Sec- inert matrix fuel in eight of the nine final cases. The MOX target
tion 4.2. The genetic algorithm varied the internal geometry of fuel does not appear in the top 30 cases in the final stage, though
these five target fuel pins, reconsidering the fuel type, the spectral both the inert matrix and hydride target fuels are well represented
shift absorber, and the outer diameters of these two materials. The in these cases (17 instances of inert matrix fuel and 13 instances of

a) Plutonium to curium transmutation ratio as a function of EFPD penalty

b) Plutonium to curium transmutation ratio as a

function of number of target fuel pins
Fig. 4.3. Plutonium to curium transmutation ratio comparisons for quarter-assemblies with one to 25 target fuel pins in the second optimization stage.
208 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212

Fig. 4.4. Quarter-assembly configuration considered in the final optimization stage.

hydride fuel). Though the final solutions are weighted towards absorber coating thicknesses are a product of the evolution process
inert matrix and hydride fuels, each of the three fuel types used by the genetic algorithm and not a result of the algorithm
occurred regularly during the evaluations (see Table 4.6). The pos- becoming trapped in a local minimum. The following section con-
sible spectral shift absorbers were also evenly represented in the siders this further.
stage three calculations (Table 4.6). The best configuration for the
transmutation of plutonium determined by these simulations, Case 4.4. Optimization validation
8 in Table 4.5, consisted of a thin (1.44 mm) target fuel pin of the
inert matrix fuel with a 0.064 mm lutetium oxide spectral shift Overarching analysis of both the single (stage one) and multiple
absorber coating. Determination of the most optimal case resulted (stage three) target fuel pin optimizations reveals no quarter-
from the comparison of the linear distance of Objective function assembly configurations that resulted in a net increase in the mass
Two (x) and Three (y) to zero for each individual case in the third of the minor actinides tracked in this study. Fig. 4.6 presents the
optimization stage. Case 8 represents the optimal case because its normalized mass of plutonium at end of cycle as a function of
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
objective functions are the closest (d ¼ x2 þ y2 ) to the ideal case the normalized mass of minor actinides at end of cycle for all of
(0, 0). the stage one and stage three cases. The total quantity of minor
The configurations in Table 4.5 result in a slight increase in the actinides decreased in all cases evaluated in this study, as seen in
transmutation of minor actinides over the irradiation period com- Fig. 4.6. Since all of the evaluations performed in this study gener-
pared to the first optimization stage (52 wt% for Case 8 in the final ated additional curium and the other residual minor actinides are
stage and 51 wt% for Case 5 in the first stage). The radius of the tar- neptunium and americium, the net decrease in minor actinides
get fuel pin increased by 240% in the final optimization stage com- must come from the transmutation of americium and neptunium.
pared to the radius of the target fuel pin in the first optimization Fig. 4.7 presents the end of life normalized masses of
stage. While the total mass of the target fuel pin remains small, plutonium-239 and curium-244 for all of the simulations per-
compared to the uranium oxide fuel rods, the improvements from formed during this study and in previous related studies
the first optimization stage are significant. Unexpectedly, the coat- (Washington et al., 2017). The masses presented in Fig. 4.7 are nor-
ing thickness of the spectral shift absorber coating did not vary in malized to the initial masses in the target fuels. The Pareto frontier
the top nine configurations, regardless of the material selected for is visible on the left edge of this figure and the masses of
the absorber coating. Though the spectral shift absorber thickness plutonium-239 and curium-244 act as competing variables. The
is uniform in the results, the simulations did explore the full prob- distribution of the data points in Fig. 4.7 is broad and indicates
lem space (see Fig. 4.5). This indicates that the final spectral shift no local maxima in which the genetic search algorithm could

Fig. 4.5. Spectral shift absorber outer radius as a function of evaluation number in the final optimization stage.
 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212 209

Table 4.5
Near optimal multiple target fuel pin designs from the stage three evaluations.

Case Fuel pin radius (mm) Absorber thickness (mm) Spectral shift absorber Fuel type Objective function 2 Objective function 3
1 2
1 2.25  10 6.41  10 Gd2O3 PuZrO2MgO (8%) 0.054068 0.531486
2 7.24  101 6.41  102 Gd2O3 PuZrO2MgO (8%) 0.058489 0.509038
3 9.37  101 6.41  102 Gd2O3 PuZrO2MgO (8%) 0.064093 0.503688
4 1.05 6.41  102 Gd2O3 PuZrO2MgO (8%) 0.067443 0.501492
5 1.01 6.41  102 Sm2O3 PuZrO2MgO (8%) 0.072549 0.493598
6 1.01 6.41  102 Lu2O3 PuZrO2MgO (8%) 0.078174 0.484607
7 1.05 6.41  102 Lu2O3 PuZrO2MgO (8%) 0.078806 0.484052
8 1.44 6.41  102 Lu2O3 PuZrO2MgO (8%) 0.086781 0.480742
9 7.24  101 6.41  102 Lu2O3 Pu0.31ZrH1.6Th1.08 0.106412 0.47843

Table 4.6
Frequency of each fuel type and spectral shift absorber during the stage three evaluations.

Fuel type Count Frequency SSA type Count Frequency

PuZrO2MgO (8%) 277 43.49% Gd2O3 169 26.53%
Pu0.31ZrH1.6Th1.08 233 36.58% In2O3 171 26.84%
MOX (5% PuO2) 127 19.94% Lu2O3 181 28.41%
Sm2O3 116 18.21%

have become trapped. The large distribution indicates that the points (diamonds) represent optimization steps that were
algorithm explored a majority of the problem space. The low quan- performed by hand using the spectral shift absorbers listed in
tity of plutonium transmutation displayed in the upper left corner Table 2.1 with all of the potential fuel types considered in the pre-
of Fig. 4.7 results from those target fuel pins being self-shielded vious study. These instances of the optimization model approached
from the neutron flux, which prevents significant burnup of pluto- the Pareto frontier developed in this study but did not fully explore
nium and also reduces the total quantity of curium produced dur- the possible solutions to this optimization problem. The first pass
ing the simulations. This self-shielding occurs when the spectral of the coupled NEWT and MOGA packages provided the triangular
shift absorber coating is too thick for depletion over the burnup data points. These data points are clustered in the upper right of
cycle of the quarter-assembly. The converse is true for curium- the plot due to an initial overweighting of the first objective func-
244 production. Higher plutonium transmutation results in the tion listed in Table 4.1. All three of these data sets provided initial
production of additional curium-244 through capture reactions. information to the final three optimization stages (squares in
In this study, the rate of change for curium-244 transmutation Fig. 4.7).
has a significantly smaller gradient than that of the plutonium- The three optimization stages in the present study developed
239 transmutation. This is visible in the lower region of Fig. 4.7. the same Pareto frontier as the previous studies (shown in
The target fuel evaluations shown as round dots in Fig. 4.7 rep- Fig. 4.7). This indicates the limiting plutonium and curium trans-
resent the end of life data for single target fuel pins of various fuel mutation rates that can be achieved using the geometries explored
types without a spectral shift absorber. These evaluations come in this study. Further improvements in relative transmutation rates
from a previous study (Washington et al., 2017) and are included would have to be achieved through additional alterations to the
to demonstrate the improvements provided by the addition of a quarter-assembly geometry (such as lattice pitch), changes to the
spectral shift absorber coating. The absorber optimization data total number of fuel pins, or increased irradiation times. The curva-

Fig. 4.6. Normalized mass of plutonium at end of cycle as a function of the normalized mass of the minor actinides at end of cycle for all of the stage one and stage three cases.
210 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212

Fig. 4.7. Normalized mass of plutonium-239 as a function of normalized mass of curium-244.

ture of the Pareto front demonstrates that achieving transmutation configurations. In each of the cases shown, the addition of the
of all of the plutonium in the target fuel pins is unfavorable in spectral shift absorber coating improves the transmutation of
terms of curium production. Transmuting more than 90 wt% of the minor actinides. The first and third optimization stages
plutonium-239 significantly increases the quantity of curium-244 rejected In2O3 as an optimal spectral shift coating despite the
produced in the target fuel pins. performance of In2O3 in the absorber optimization evaluations
(Washington et al., 2017). The frequency of the In2O3 absorber
4.5. Optimization results (Tables 4.3 and 4.6) indicates it was well represented during
the optimization stages but less effective than the other
Though this study did not indicate the possibility of complete spectral shift coatings in reducing the objective function values.
transmutation of all the minor actinides and plutonium, a clear In Cases 1 through 5 in Table 4.7, the addition of the coating
balance between plutonium transmuted and curium produced also improves the total plutonium transmutation. The
exists. The curvature of the Pareto front seen in Fig. 4.7 shows an highest transmutation of plutonium and minor actinides was
optimum trade-off of plutonium-239 transmutation and curium- achieved through the irradiation of a PuZrO2MgO target
244 production between a 70–90 wt% reduction in plutonium- fuel with a Lu2O3 spectral shift absorber coating, Case 8 in
239 and a 250–400% increase in curium-244. All of the simulations Table 4.7.
performed during this study resulted in a reduction in the total The optimal target fuel pins selected by the present study are
quantities of minor actinides (see Fig. 4.6). very thin. The production of an actual fuel pellet with a radius of
The addition of the spectral shift absorber coating to the tar- 0.225 mm (Case 1 in Table 4.7), for example, would be very diffi-
get fuel pins provides an increase in the transmutation of the cult. The small radial dimension of the fuel reduces the self-
minor actinides at the cost of a decrease in the total rate of plu- shielding of the material and has additional implications assuming
tonium transmutation. These differences are shown in Table 4.7 the fuel pins could be produced. For example, the small quantity of
as the percent change in the objective functions due to the fissile material contained in the fuel pin could warrant a reduction
addition of a spectral shift absorber to the top nine stage-three in the cladding thickness of the fuel pin. Pins of this size could be

Table 4.7
Difference in objective functions for the nine optimal stage three configurations from DAKOTA with and without spectral shift absorbers.

Case Fuel pin Absorber Spectral shift Fuel type Change in objective Change in objective
radius (mm) thickness (mm) absorber function 2a function 3a
1 2.25  101 6.41  102 Gd2O3 PuZrO2MgO (8%) 11.43% 9.59%
2 7.24  101 6.41  102 Gd2O3 PuZrO2MgO (8%) 21.12% 8.04%
3 9.37  101 6.41  102 Gd2O3 PuZrO2MgO (8%) 16.61% 7.44%
4 1.05 6.41  102 Gd2O3 PuZrO2MgO (8%) 13.71% 7.12%
5 1.01 6.41  102 Sm2O3 PuZrO2MgO (8%) 4.90% 9.07%
6 1.01 6.41  102 Lu2O3 PuZrO2MgO (8%) 2.65% 11.09%
7 1.05 6.41  102 Lu2O3 PuZrO2MgO (8%) 2.69% 10.98%
8 1.44 6.41  102 Lu2O3 PuZrO2MgO (8%) 3.24% 9.64%
9 7.24  101 6.41  102 Lu2O3 Pu0.31ZrH1.6Th1.08 1.66% 5.30%
a
(Objective function with SSA – Objective function without SSA)/Objective function without SSA.
 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212 211

incorporated into additional inert matrix materials or, depending a 1.44 mm radius and a 0.06 mm Lu2O3 coating in a five target fuel
on the ductility of the material, could be drawn into wires and pin per quarter-assembly configuration represents the near opti-
wrapped around existing fuel pellets. The radiological hazard of mal combination for the transmutation of plutonium and minor
fabrication of such small quantities of radioactive material would actinides in the LWR environment considered in this study.
be less than that of a more standard fuel pellet size though the
overall rate of transmutation would be small. Acknowledgements

Jeremy Washington was supported by a Colorado School of

5. Summary and conclusions Mines Nuclear Science and Engineering Fellowship with funding
provided by the U.S. Nuclear Regulatory Commission under grant
This study evaluated combinations of three different target number NRC-38-10-952.
fuels (MOX (5% PuO2), Pu0.31ZrH1.6Th1.08, and PuZrO2MgO) and four
different spectral shift absorbers (Gd2O3, In2O3, Lu2O3, and Sm2O3) References
for the transmutation of neptunium, plutonium, americium, and
curium in existing light water reactors as a method to reduce the Adams, B.M., Ebeida, M., Eldred, M., Jakeman, J., Swiler, L., Bohnhoff, W., Dalbey, K.,
stockpile of plutonium and minor actinides without the use of fast Eddy, J., Hu, K., Vigil, D., Bauman, L., Hough, P., 2013. Dakota, A Multilevel
Parallel Object-Oriented Framework for Design Optimization, Parameter
reactors. Coupled NEWT/DAKOTA models performed the optimiza- Estimation, Uncertainty Quantification, and Sensitivity Analysis Version 5.4
tion studies in this study. The Multi-Objective function Genetic Theory Manual, 5.4 ed., Sandia National Laboratories.
search Algorithm (MOGA) module of the DAKOTA optimization Aghaie, M., Norouzi, A., Zolfaghari, A., Minuchehr, A., Mohamadi Fard, Z., Tumari, R.,
2013. Advanced progressive real coded genetic algorithm for nuclear system
suite provided the genetic search algorithm. The NEWT package availability optimization through preventive maintenance scheduling. Ann.
of the SCALE reactor code performed the neutronics calculations, Nucl. Energy 60, 64–72. http://dx.doi.org/10.1016/j.anucene.2013.04.023.
based on a modified AP1000 17  17 Robust Fuel Assembly geom- Alim, F., Ivanov, K., Levine, S.H., 2008. New genetic algorithms (GA) to optimize PWR
reactors Part I: loading pattern and burnable poison placement optimization
etry. The combinations of target fuel and spectral shift absorber techniques for PWRs. Ann. Nucl. Energy 35, 93–112.
were evaluated for the destruction of plutonium and minor acti- Baraldi, P., Pedroni, N., Zio, E., 2009. Application of a niched Pareto genetic algorithm
nides as a function of target fuel pin material, spectral shift absor- for selecting features for nuclear transients classification. Int. J. Intell. Syst. 24,
118–151. http://dx.doi.org/10.1002/int.20328.
ber material, internal target fuel pin geometry, placement of target Chadwick, M.B., Oblozinsky, P., Herman, M., Greene, N.M., McKnight, R.D., Smith, D.
fuel pins in the quarter-assembly, and number of target fuel pins. L., Young, P.G., MacFarlane, R.E., Hale, G.M., Frankle, S.C., Kahler, A.C., Kawano,
Normalized objective functions compared each configuration at T., Little, R.C., Madland, D.G., Moller, P., Mosteller, R.D., Page, P.R., Talou, P.,
Trellue, H., White, M.C., Wilson, W.B., Arcilla, R., Dunford, C.L., Mughabghab, S.F.,
the end of a 1400 EFPD burnup cycle.
Pritychenko, B., Rochman, D., Sonzogni, A.A., Lubitz, C.R., Trumbull, T.H.,
The optimization proceeded in three stages. The first stage opti- Weinman, J.P., Brown, D.A., Cullen, D.E., Heinrichs, D.P., McNabb, D.P., Derrien,
mized a single target fuel pin per quarter-assembly located adja- H., Dunn, M.E., Larson, N.M., Leal, L.C., Carlson, A.D., Block, R.C., Briggs, J.B.,
cent to the central instrumentation channel. This stage revealed Cheng, E.T., Huria, H.C., Zerkle, M.L., Kozier, K.S., Courcelle, A., Pronyaev, V., van
der Marck, S.C., 2006. ENDF/B-VII.0: next generation evaluated nuclear data
that a thin (0.036 mm) inert matrix target fuel pin (PuZrO2MgO) library for nuclear science and technology. Nucl. Data Sheets 107, 2931–3060.
with a lutetium oxide (Lu2O3) coating resulted in the most com- Department of Energy, 2002. Environmental Impact Statement for a Geologic
plete transmutation (88 wt% of the plutonium and 51 wt% of the Repository for the Disposal of Spent Nuclear Fuel and High-Level Radioactive
Waste at Yucca Mountain, Nye County, Nevada (DOE/EIS-0250). U.S.
minor actinides). Department of Energy.
In the second optimization stage, the coupled codes evaluated Duderstadt, J.J., Hamilton, L.J., 1976. Nuclear Reactor Analysis. John Wiley & Sons.
quarter-assembly configurations with different numbers of the Inc., New York, NY.
Dunford, C.L., Kinsey, R.R., 1998. NuDat system for access to nuclear data,
optimal first stage target fuel pins in various locations in the International Atomic Energy Agency report, IAEA-NDS-205 (BNL-NCS-65687).
quarter-assembly. Though the highest rated quarter-assemblies International Atomic Energy Agency, Vienna, Austria.
contained three or fewer target fuel pins, a number of data points Eddy, J., Lewis, K., 2001. Effective generation of Pareto sets using genetic
programming. In: Proceedings of ASME Design Engineering Technical
suggested additional target pins could be added. The second stage Conference.
optimization indicated that the highest rate of minor actinide Galperin, A., 1995. Utilization of light water reactors for plutonium incineration.
transmutation would occur in a quarter-assembly containing five Ann. Nucl. Energy 22, 507–511.
Gen, M., Cheng, R., Lin, L., 2008. Network Models and Optimization. Springer-Verlag,
target fuel pins. Irradiation of the five target pin quarter-
London Limited.
assembly selected for further optimization achieved transmutation Jiejuan, T., Dingyuan, M., Dazhi, X., 2004. A genetic algorithm solution for a nuclear
rates similar to some of the one and two target fuel pin quarter- power plant risk–cost maintenance model. Nucl. Eng. Des. 229, 81–89. http://
assemblies. The selected five pin quarter-assembly also ended dx.doi.org/10.1016/S0029-5493(03)00210-3.
Lide, D.R., 2009. CRC Handbook of Chemistry and Physics. Boca Raton, FL.
the irradiation period with a relatively small EFPD penalty (75 Norouzi, A., Zolfaghari, A., Minuchehr, A.H., Khoshahval, F., 2011. An enhanced
EFPD). integer coded genetic algorithm to optimize PWRs. Prog. Nucl. Energy 53, 449–
The third optimization stage re-optimized the geometry of the 456.
Oak Ridge National Laboratory, 2011. SCALE: A comprehensive modeling and
target fuel pins in the five target pin quarter-assembly geometry simulation suite for nuclear safety analysis and design.
from the second optimization stage to maximize plutonium and Olander, D., Greenspan, E., Garkisch, H.D., Petrovic, B., 2009. Uranium–zirconium
minor actinide transmutation. This stage fixed the number hydride fuel properties. Nucl. Eng. Des. 239, 1406–1424.
Özdemir, L., Acar, B.B., Zabunoğlu, O.H., 2011. Determination of fissile fraction in
and position of the five target fuel pins in the quarter assembly MOX (mixed U + Pu oxides) fuels for different burnup values. Ann. Nucl. Energy
and re-optimized the pins’ internal geometry. The final stage pro- 38, 540–546.
duced a quarter-assembly capable of transmuting nearly 89 wt% Pereira, C.M., Lapa, C.M., 2003. Coarse-grained parallel genetic algorithm applied to
a nuclear reactor core design optimization problem. Ann. Nucl. Energy 30, 555–
of the original plutonium and 52 wt% of the original minor acti- 565.
nides by the end of an irradiation cycle. The third stage configura- Pereira, C.M.N.A., Sacco, W.F., 2008. A parallel genetic algorithm with niching
tion resulted in larger target fuel pins than the first stage (fuel pin technique applied to a nuclear reactor core design optimization problem. Prog.
Nucl. Energy 50, 740–746. http://dx.doi.org/10.1016/j.pnucene.2007.12.007.
radii of 1.44 mm vs 0.6 mm, respectively) with a thinner coating of
Salvatores, M., Palmiotti, G., 2011. Radioactive waste partitioning and
spectral shift absorber (coating thickness of 0.06 mm vs 0.85 mm, transmutation within advanced fuel cycles: achievements and challenges.
respectively). Additionally, the variety of potentially optimal spec- Prog. Part. Nucl. Phys. 66, 144–166.
tral shift absorbers increased in the final stage, from one material Schulz, T.L., 2006. Westinghouse AP1000 advanced passive plant. Nucl. Eng. Des.
236, 1547–1557.
(Lu2O3) in the first stage to three (Gd2O3, Lu2O3, and Sm2O3) in Trellue, H.R., 2006. Safety and neutronics: a comparison of MOX vs UO2 fuel. Prog.
the final stage. The 8 wt% PuZrO2MgO inert matrix fuel pin with Nucl. Energy 48, 135–145.
212 J. Washington, J. King / Nuclear Engineering and Design 311 (2017) 199–212

Washington, J., King, J., Shayer, Z., 2017. Target fuels for plutonium and minor Yilmaz, S., Ivanov, K., Levine, S., Mahgerefteh, M., 2006a. Application of genetic
actinide transmutation in pressurized water reactors. Nucl. Eng. Des. 313, algorithms to optimize burnable poison placement in pressurized water
53–72. reactors. Ann. Nucl. Energy 33, 446–456.
Washington, J., King, J., Shayer, Z., 2014. Selection and evaluation of potential Yilmaz, S., Ivanov, K., Levine, S., Mahgerefteh, M., 2006b. Development of enriched
burnable absorbers incorporated into modified TRISO particles. Nucl. Eng. Des. Gd-155 and Gd-157 burnable poison designs for a PWR core. Ann. Nucl. Energy
278, 377–386. http://dx.doi.org/10.1016/j.nucengdes.2014.07.033. 33, 439–445.
Westinghouse Electric Corp, Foster, J.P., Comstock, R.J., Worcester, S.A., Sabol, G.P., Zhao, J., Knight, B., Nissan, E., Soper, A., 1998. FUELGEN: a genetic algorithm-based
1992. ZIRLO material for light water reactor applications. 5112573. system for fuel loading pattern design in nuclear power reactors. Expert Syst.
Wigeland, R.A., Bauer, T.H., Fanning, T.H., Morris, E.E., 2006. Separations and Appl. 14, 461–470.
transmutation criteria to improve utilization of a geologic repository. Nucl.
Technol. 154, 95–106.