The Advanced Fuel Cycle Initiative Overview of NEAMS Reactor Simulation Project Andrew Siegel Argonne National Laboratory NEAMS Reactor Meeting May 19,

The Advanced Fuel Cycle Initiative

Overview of NEAMS Reactor Simulation Project

Andrew Siegel

Argonne National Laboratory

NEAMS Reactor Meeting

May 19, 2009

NEAMS Reactor Workshop

NEAMS Campaign

20M total for FY09

Fast Reactor Simulation Component (~5.5M)

– Advanced Neutronics Modeling (ANL/INL): 3-4 FTEs

– Advanced Thermal Hydraulics Modeling (ANL/UIUC): 3-4 FTEs

– Advanced Framework Design (ANL/LLNL/UW): 3-4 FTEs

– Advanced Safety Modeling (ANL):1-2 FTEs

Other Program components

– IPSCs: Fuels (~6.5M), Separations (500K), waste forms (1M?)

– Cross cutting: V&V (2M), FMM (1M), ECT (500K), CT (500K)

NEAMS Reactor Workshop

Vision Develop process/methodologies to enable the use of computer

simulation in a fundamentally new way for operation, design, and licensing of nuclear systems.

Treat simulations as numerical experiments– Model physics instead of devices– Solve governing equations of motion on detailed 3-D grids– Do simulations prior to physical experiments --> downselect– Leverage massive computing power (petascale) + HPC expertise– Combine single-effects validation to infer behavior of integrated system

Goals– Reduce operating margins for existing designs– Explore innovative designs with minimal reliance on mock-up

experiments– Explore innovative materials for new and existing designs– Reduce regulatory cost

JFR DesignConcepts

NEAMS Reactor Workshop

Key physics models for fast reactor

Non-boiling Thermal Hydraulics

– Equations well known

– Some degree of subgrid-scale modeling generally required because of computational limitations (Re ~50K per channel)

• Many available SGS -- what is good enough for problems of interest?

• Most not developed for low-Pr flows

– Meshing issues

– Improving scalability to leverage leadership class machines

– Application to specific problems: understanding sensitivities for fast-reactor specific problems

– Embedded sensitivity analysis challenging

– initial problems: rod bundle mixing, pressure drop, upper plenum mixing, shut down heat removal

NEAMS Reactor Workshop

Key physics for fast reactor, cont.

Neutron transport

– Equations well known

– Very mature field

– Non-homogenized deterministic solutions not possible due to computing limitations

– Monte Carlo simulations still not practical for most design studies because of computing limitations

– Reducing/eliminating homogenization important for high-fidelity coupling, reducing uncertainties (particularly for new designs)

– When does data uncertainty swamp method uncertainty?

– Work on efficient algorithms, highly scalable algorithms, application to benchmark problems

NEAMS Reactor Workshop

Key physics for fast reactor, cont.

Structural mechanics

– Critical for accurate modeling of fast reactor passive safety

– Constitutive models/basic properties exist for well known materials in ideal conditions

– Potential to use lower-length-scale modeling to derive better constitutive relations

– Long-term material strength irradiation damage/thermal load issues that can potentially be studied with more fundamental modeling

Fuel

– Dynamical equations not well known

– From reactor standpoint critical to characterize properties accurately

– Legacy codes (LIFE-Metal) ok for near term

– Strong interaction with fuels modeling team necessary longer-term

NEAMS Reactor Workshop7

445TF/s / 557 TF/s64 TB / 80TB

BG/P SupercomputerAt Argonne

Top 500 List of Supercomputers

NEAMS Reactor Workshop

Integrated Performance and Safety Code

Ultimately must be capable of performing all necessary design and safety calculations -- massive effort

– Safety: DBA, BDBA (ULOHS, UTOP, ULOF)– Includes entire plant -- core, vessel, pumps, heat exchangers, …– Design: fuel cycle analysis, hot channel analysis, peak temperature, etc.

First principles calculation impossible for full-device simulations

Multi-scale approach uses detailed local simulations to derive specific parameterizations for full-device models.

Also study specific local phenomena by embedding high-fidelity solver in coarser calculation

Start by advanced methods development, validation/verification and exploration

9Work conducted byANL for the GNEP

Spatial domain (mesh) hub for coupling– Bulk of data there– Performance most sensitive there

Function interfaces (APIs), not data structures– Fewer code dependencies

Simplify connection of components, to allow horizontal, vertical integration– Take advantage of Science improvements– Components don't need to cede control– Flexibility to take best tool for each task– Finer-grained pieces give more options

Framework philosophy

Nekton: ANL Advanced Reactor Thermal Hydraulics Code

May 2, 2009 JAEA Visit 11

Current development path– Highly-scalable high-order

LES/DNS • Supplemented initially with

commercial RANS, legacy lower resolution models

– Fast-running low resolution methods (sub-channel models)

• To provide rapid turn around for engineering design

– Highly-scalable high-order RANS

– RANS-informed lower resolution models

– LES-informed RANS models

– DNS-informed LES models

Lower Resolution Models(~750 points/channel)

Multi-Resolution Thermal Hydraulic Simulation Hierarchy

BoundaryConditions

ModelingParameters

Reynolds Averaged Navier Stokes(~20 K point/channel)

BoundaryConditions

ModelingParameters

Large Eddy Simulation(~5 M points/channel)

BoundaryConditions

ModelingParameters

Direct Numerical Simulation(~50 M points/channel)

Incr

easi

ng

Res

olu

tion

Increasin

g Dom

ain S

izeMulti-Resolution Approach

UNIC: ANL Advanced Transport Solver

May 2, 2009 JAEA Visit 13

Adaptive Flux Solution Options

Homogenized assembly

Homogenized assembly internals

Homogenized pin cells

Fully explicit assembly

Unified geometrical framework

– Unstructured finite element analysis for coupling with structural mechanics and thermal-hydraulics codes

Safety Modeling: Integrating high-fidelity models with SAS4A/SASSYS-1

15

Comparison Between RANS and Subchannel Results

Axially-independent cross flow terms used in the subchannel model are not able to resolve the axial periodicity in the temperature due to the wire wraps (see arrows).

Temperature distribution is symmetric in the subchannel results, but skewed in the RANS results. (Unanticipated bias)

Cross flow terms from higher-fidelity modeling would result in better agreement between subchannel and RANS.

Differences Between Steady-State Subchannel and RANS Coolant Temperature Distributions in a 217-Pin Fuel Bundle.