Sie sind auf Seite 1von 25

Empirical

p Diffusion
Coefficients for Natural-
U
Uranium
i CANDU Lattices
L tti
A. Patel & E. Nichita
Faculty of Energy Systems and Nuclear Science
University of Ontario Institute of Technology

ANS Annual Meeting


2010-06-15
San Diego,
g CA
Outline
„ Introduction
„ Objective
„ Approach
pp
„ Calculations
„ Results
„ Conclusion
„ Future Investigations

Outline
2
Full Core Calculations
„ Successive approximations
1. many-group heterogeneous transport (reference)
2. two-group node-homogenized transport
3. two-group
t node-homogenized
d h i d diff
diffusion
i
„ Observation: The difference between step 1 and
stepp 2 results is ((much)) smaller than the difference
between step 2 and step 3 results.
„ Initial Objective: Adjust diffusion coefficient so that
t
two-group node-homogenized
d h i d diff
diffusion
i resultslt
closely match two-group node-homogenized
transport results for CANDU lattices.

Introduction
3
Broader Objective
„ Develop equivalence between node-
node
homogenized two-group transport model and
two-group diffusion-like model, such that the
node-integrated reaction rates calculated
using the two models are the same.

Objective
4
Angle-Integrated Neutron
Balance Equation
„ Transport
G tr G G tr G G tr G 1 G tr G
∇ ⋅ J g (r ) + Σrg (r )Φ g (r ) − ∑Σ g '→g (r )Φg ' (r ) = tr χ g ∑νΣ fg' (r )Φg ' (r )
g '≠ g keff g'

„ Diffusion

[ G G
] G G G G
− ∇ ⋅ Dg (r )∇Φ g (r ) + Σrg (r )Φ g (r ) − ∑Σ g '→g (r )Φg ' (r ) =
1
keff
G G
χ g ∑νΣ fg' (r )Φg ' (r )
g '≠ g g'

Approach
5
Comments
„ Adjusting diffusion coefficients to achieve
direct equivalence between transport and
diffusion in 3D requires directional diffusion
coefficients.
„ Equivalence is much easier to achieve when
using a simplified diffusion equation:
G G G G G G 1 G G
− Dg (r )∇2Φg (r ) + Σrg (r )Φg (r ) − ∑Σ g '→g (r )Φg ' (r ) = χ g ∑νΣ fg' (r )Φg ' (r )
g '≠ g keff g'

Approach
6
New Problem
„ Find equivalence between simplified diffusion
and transport in two energy groups for CANDU
lattices.
G tr G G tr G G tr G 1 G tr G
∇ ⋅ J g (r ) + Σrg (r )Φg (r ) − ∑Σ g '→g (r )Φg ' (r ) = tr χ g ∑νΣ fg' (r )Φg ' (r )
g '≠ g keff g'

G G G G G G 1 G G
− Dg (r )∇2Φg (r ) + Σrg (r )Φg (r ) − ∑Σ g '→g (r )Φg ' (r ) = χ g ∑νΣ fg' (r )Φg ' (r )
g '≠ g keffff g'

Approach
7
Equivalence
„ Transport Equation
G tr G G tr G G tr G 1 G tr G
∇ ⋅ J g (r ) + Σrg (r )Φ g (r ) − ∑Σ g '→g (r )Φg ' (r ) = tr χ g ∑νΣ fg' (r )Φg ' (r )
g '≠ g keff g'

G tr G
G ∇ ⋅ J g (r ) G 2 tr G G tr G
„ Define: Dg (r ) = − 2 tr G ⇒ −Dg (r )∇ Φg (r ) = ∇ ⋅ J g (r )
∇ Φg (r )

„ Transport flux satisfies simplified


simplified-diffusion
diffusion eq
eq.
G G G G G G 1 G G
− Dg (r )∇2Φtrg (r ) + Σrg (r )Φtrg (r ) − ∑Σ g '→g (r )Φtrg' (r ) = χ g ∑νΣ fg' (r )Φtrg' (r )
g '≠ g keffff g'

Approach
8
Comments
„ If the transport flux is known, a simplified
diffusion problem can be set up whose
solution produces the same integral flux as
the transport problem.
„ Possible difficulties setting up the simplified
diffusion problem:
„ Transport flux not known beforehand
„ Transport flux may be known in a form that is
not amenable to calculation of Laplacian (e.g.
irregular-region CP)
Approach
9
CANDU-Problem-Specific
Features
„ Equivalence is sought between two node
node-
homogenized models.
„ For homogenized
g nodes a simple
p Cartesian
mesh can be used, which allows easy
computation of Laplacian.
„ Homogenized diffusion coefficients in CANDU
lattices vary little with local conditions and
h
hence could
ld b
be approximated
i d to b
be constant
in space.

Approach
10
Simplification
„ Assumption
Diffusion coefficients for the simplified diffusion
„
equation are almost constant throughout the system
and have very similar values regardless of the system
configuration.
„ Consequences
„ Using constant (average) diffusion coefficients
throughout a system yields almost the same results as
using the position-dependent ones.
„ Average
A simplified-diffusion
i lifi d diff i coefficients
ffi i t ((one value
l ffor
each group) can be calculated for a simple
configuration and then applied to any other
configuration.
fi ti
Approach
11
Determining the Empirical
Diffusion Coefficients
„ For a simple configuration, calculate:
G G tr G 3
∫ wg (r )∇ ⋅ J g (r )d r G tr G
G ∇ ⋅ J g (r )
Dg = ≅ Dg (r ) = 2 trt G
G 2 tr G 3 ∇ Φg (r )
∫ g )∇ Φg (r )d r
w (r

„ If currents are not directly available


available, use:
G⎡ G tr G G tr G ⎤ 3 1 G G tr G 3
∫ g ⎢⎣ rg g ∑
w ( r ) Σ ( r )Φ ( r ) −
g '≠
Σ g '→g ( r )Φ g' ( r ) ⎥

d r −
k tr ∫ g
w ( r ) χ g∑
g'
νΣ fg ' ( r )Φ g ' (r )d r
Dg =
eff

G 2 tr G 3
∫ g )∇ Φ g (r )d r
w ( r

Approach
12
Configurations (1-D)
(1 D)
„ For calculating D g
„ 11 identical homogeneous-nodes
((corresponding
g to mid-burnup fuel))
REFL F F F F VAC

„ For testing
g
„ 13 homogeneous nodes
(11 fuel + 2 reflector)
REFL F F F F R R VAC

REFL F F F F R R VAC

Calculations
13
All Mid-Burnup Fuel + Reflector
Fast Flux
1.80 15.0%
1.60 10.0%
1.40 5.0%
1.20 0.0%
1.00 ‐5.0%
0.80 ‐10.0%
0.60 ‐15.0%
0 40
0.40 20 0%
‐20.0%
0.20 ‐25.0%
0.00 ‐30.0%
1 2 3 4 5 6 7 8 9 10 11 12 13
Region

phi‐1 ‐ transport phi‐1 ‐ theoretical D


phi‐1 ‐ empirical D phi‐1 % err ‐ theoretical D
phi‐1 % err ‐ empirical D

Results
14
All Mid-Burnup Fuel + Reflector
Thermal Flux
3.50 12.0%

3.00 10.0%

2.50 8.0%

2 00
2.00 6 0%
6.0%

1.50 4.0%

1.00 2.0%

0.50 0.0%

0.00 ‐2.0%
1 2 3 4 5 6 7 8 9 10 11 12 13
Region

phi‐2 ‐ transport phi‐2 ‐ theoretical D


phi‐2 ‐ empirical D phi‐2 % err ‐ theoretical D
phi‐2 % err ‐ empirical D

Results
15
All Mid-Burnup Fuel + Reflector
Power
1.60 2.5%

1.40 2.0%

1.20 1.5%

1 00
1.00 1 0%
1.0%

0.80 0.5%

0.60 0.0%

0.40 ‐0.5%

0.20 ‐1.0%

0.00 ‐1.5%
1 2 3 4 5 6 7 8 9 10 11
Region

power ‐ transport power ‐ theoretical D power ‐ empirical D


% error ‐ theoretical D % error ‐ empirical D

Results
16
All Mid-Burnup Fuel + Reflector
Error Synopsis
D calculation method
theoretical empirical
keff error ((mk)
k) 1.6
1 6 -0.1
01
max 19.2% 4.6%
Φ1 error
RMS 5 4%
5.4% 1 4%
1.4%
max 9.4% 1.3%
Φ2 error RMS 3.3%
3 3% 0.6%
0 6%
max 2.2% 1.3%
power error
RMS 0.7% 0.5%

Results
17
Disch. Fuel + Fresh Fuel + Refl.
Fast Flux
3.00 40.0%

2.50 30.0%

2.00 20.0%

1.50 10.0%

1.00 0.0%

0.50 ‐10.0%

0.00 ‐20.0%
1 2 3 4 5 6 7 8 9 10 11 12 13
Region
phi‐1 ‐ transport phi‐1 ‐ theoretical D
phi‐1 ‐ empirical D phi‐1 % err ‐ theoretical D
phi‐1 % err ‐ empirical D
p p

Results
18
Disch. Fuel + Fresh Fuel + Refl.
Thermal Flux
3.00 40.0%

2.50 30.0%

2.00 20.0%

1.50 10.0%

1.00 0.0%

0.50 ‐10.0%

0.00 ‐20.0%
1 2 3 4 5 6 7 8 9 10 11 12 13
Region
phi‐2 ‐ transport phi‐2 ‐ theoretical D
phi‐2 ‐ empirical D phi‐2 % err ‐ theoretical D
phi‐2 % err ‐ empirical D
p p

Results
19
Disch. Fuel + Fresh Fuel + Refl.
Power
1.60 25.0%

1.40 20.0%

1.20 15.0%

1 00
1.00 10 0%
10.0%

0.80 5.0%

0.60 0.0%

0.40 ‐5.0%

0.20 ‐10.0%

0.00 ‐15.0%
1 2 3 4 5 6 7 8 9 10 11
Region
power ‐ transport power ‐ theoretical D
power ‐ empirical D % error ‐ theoretical D
% error ‐ empirical D

Results
20
Disch. Fuel + Fresh Fuel + Refl.
Error Synopsis
D calculation method
theoretical empirical
keff error ((mk)
k) 2.5
2 5 -0.2
02
max 22.9% 3.4%
Φ1 error
RMS 10 8%
10.8% 1 4%
1.4%
max 33.5% 2.4%
Φ2 error RMS 15.9% 1.3%
max 23.8% 2.4%
power error
RMS 10.5% 1.1%

Results
21
Comments
„ Empirical diffusion coefficients allow the
simplified diffusion results to match closely
transport results.
„ Because the diffusion coefficients are
assumed constant in space, existing codes
do not need to be modified to implement the
simplified diffusion equation.

Results
22
Conclusion
„ Using empirical diffusion coefficients for
CANDU-lattice node-homogenized models
yields substantial gains in accuracy for simple,
one-dimensional, configurations.

Conclusion
23
Future Investigations
„ Confirm improvement
p in accuracyy occurs for two-
and three-dimensional models.
„ Develop an interpretation of the difference
between theoretical and empirical diffusion
coefficients.

Future Investigations
24
Questions

Questions
25

Das könnte Ihnen auch gefallen