Advanced Neutronics: PHISICS project

hesitantdoubtfulΤεχνίτη Νοημοσύνη και Ρομποτική

29 Οκτ 2013 (πριν από 3 χρόνια και 7 μήνες)

68 εμφανίσεις

Advanced Neutronics:
PHISICS project

C. Rabiti, Y. Wang, G. Palmiotti,
A. Epiney, A. Alfonsi, H. Hiruta,
J. Cogliati, T. Grimmett

What is PHISICS?

P
arallel and
H
ighly
I
nnovative
S
imulation for
I
NL
C
ode
S
ystem


PHISICS was started two years ago


The idea behind the project is to provide state of the art
simulation capability to reactor physics designers


Key features are


Modeling flexibility

Hardware/software flexibility

Long term maintainability

Uncertainty analysis

Focuses


Neutronic design

Fluxes, burn up, fuel cycle

Core feedback to system codes

PHISICS & RELAP5 !?!?


PHISCS was born under one constrain:

increasing accuracy should not come at the expense of user
time


How?


Take advantage of
parallel computing

available to
the
average user


Introduce a
more complete representation

of the
physical phenomena

These are the reasons why we believe PHISICS is
a good candidate to move RELAP5 forward

PHISICS Structure


Modular infrastructure
to ensure easy upgrade of
components and maintenance. No deep
interdependence of modules





Kernel

Interface

Kernel

Interface

Data
Type

Data
Type

Data
Type

Data
Type

Input

Input

Driver

PHISICS
-
RELAP5: Coupling

RELAP 5:

Plant and TH

XS
-
MIXER

INSTANT

T
f
,
T
c
,
ρ
c
,…

XS

Power

Bateman

Solver

Flux

Nuclide Densities

Fission Power

Time Driver

Decay heat

XS

Source

Following the arrow clockwise, the loop reproduces an
operator split approach

Overview of the Components


INSTANT: Transport/Diffusion Solver



MRTAU: Bateman Solver



Adjoint Perturbation Theory Module



Time Dependent Driver

INSTANT

Intelligent Nodal and
Semistrucured

Treatment for Advanced
Neutron Transport



INSTANT is in continuous
development to extend its
capability



Code is designed to take full
advantage of middle to large
cluster (
10~1000 processors
)



Code is designed to focus on
method adaptation

while also
mesh adaptation
will be possible

INSTANT: General Features


Boundary condition


Reflective (exact)

Vacuum


Periodic


Number of energy group:
unlimited

(memory)


Anisotropic order of the scattering: up to P33


Number of thermal groups (up
-
scattering):
unlimited


Problem type:


fundamental mode,
forward

and
adjoint


Source,
forward

and
adjoint


Outer iteration: power iteration scheme accelerated with
Chebyshev


Inner iteration acceleration: diffusion partitioning


INSTANT: Nodal Mesh/Geometry


INSTANT has the capability to treat (with a nodal approach) the
following geometry/mesh capability


2/3D Cartesian


2D triangular, Z extruded


2D hexagonal, Z extruded

Takeda 1 Benchmark (Rod In)

PN order

Keff

1 (diffusion)

0.93238

3

0.96117

5

0.96214

7

0.96230

Takeda 1 Benchmark (Rod In)

MHTGR 2D Triangular Geometry

Unstructured Mesh


C5G7

INSTANT: Inner Iteration Algorithms


INSTANT has 3 different inner iteration schemes:









The Krylov space based solvers (CG, and GMRES), since
based on a residual formulation, are suitable for multi
-
physics
coupling using Jacobian Free Newton Krylov methods


Multi
-
color by axial
layers

Conjugate Gradient

Generalized Minimal
RESidual



Speed



Memory



Scalability

Parallel Implementation of INSTNAT

Number of
processors

Total computing
time (s)

Speed up

1

35.70

1.00

2

17.98

1.99

3

11.11

3.21

4

8.39

4.26

6

5.93

6.02

9

4.03

8.86


36x36 cells with quadratic shape functions used to discretize


the problem


Computing time on a single desktop with P9

SN
-
PN Coupling


One of the middle term goals of
INSTANT is to provide the
capability to use different
algorithms and mesh structures in
the same geometry



The theoretical developments
needed are under development in
cooperation with Texas A&M
University



Implementation in the code will
start in FY12

Perturbation Module

The tasks of the perturbation module are


Compute the uncertainty due to cross section in


Keff


Reactivity feedbacks


Reaction rates


Requirements


Classical perturbation theory (adjoint in fundamental mode)


Generalized Perturbation Theory (source adjoint in critical
problems)


Cross section manipulation tools


Perturbation integral computation

Adjoint Capability


The adjoint solution has been already implemented in
INSTANT and a perturbation module is under development


the perturbation module will provide sensitivity analysis of
input parameters (cross sections) toward several figure of
merit:


Keff


Power peak


Control rod worth


Reaction rates


Etc.


Unstructured Mesh Adjoint

MRTAU: Deacy/Depletion

M
ulti
-
R
eactor
T
ransmutation
A
nalysis
U
tility


This module solves the Bateman equation


Two solution methodologies are available


Arbitrary order of Taylor development


Arbitrary Chebyshev Rational Approximation
Method order (CRAM)


CRAM is the default methodology


Comparison of CRAM and Taylor have been
performed to confirm the correct implementation of
algorithms

MRTAU Test Case


MOX (10% NpPu/HM, 50 MWD/kg) directly recycled from spent UO2
(4.2% 235U/U, 50 MWD/kg, 5 years decay storage


CRAM 14
th

order


Taylor 2
nd

order


Depletion Burn
-
Up Sequence


100 days cool down


100 days core


100 days cool down


Reference CRAM at 3000 time step (TS)


Results for few representative isotopes

100 Days Error (%)

Out of range

200 Days Error (%)

Out of range

300 Days Error (%)

Out of range

MRTAU remarks


The two implemented algorithms offers the needed
flexibility to model either burn up and decay



Study are on going to assess best trade of between
computational time and accuracy for transient
analysis using different methodologies and orders
of accuracy

Time Driver


The module is at the design stage


The algorithm is based on the same scheme
implemented in ERANOS/KIN3D


Adaptation of the original algorithms are under
study to increase speed for very short time step



We promised it by the end of September

Conclusion

PHISICS will bring a new set of capabilities to RELAP5


Pre depletion of the core


Multi group in energy


Cross section tabulation


More flexibility in the neutronic mesh


Transport


Exact decay heat and isotopic tracking


Our goal is to deliver higher accuracy without
impacting usability and computational time


Seeking Your Input

Please let us know….


what more you would like to add from the
neutronics standpoint



Why you think what we are doing is or is not
beneficial to your work


Question?