Advanced Thermal Hydraulics Simulation

1
Advanced Thermal Hydraulics Simulation Part II

• David Pointer Carlos Pantano - UIUC
• Jeff Smith Hank Childs LLNL
• Adrian Tentner
• Paul Fischer
• James Lottes
• Aleks Obabko
• Yulia Peet
• Andrew Siegel
• - Argonne National Laboratory

2
RANS-based simulations

• Commercial CFD codes STAR-CCM and STAR-CD
• Finite volume solution of the well-known
  Reynolds-averaged form of the Navier-Stokes
  equations.

• Second order solution using
• Face flux based differencing (COMET)
• SIMPLE predictor-corrector solution algorithm
• Algebraic multi-grid pre-conditioning
• two-equation realizable k-epsilon turbulence
  model

3
Allowed Mesh Types in Star-CCM

• Radial cross-sections showing computational mesh
  distributions in the 7-pin RANS simulations for
• (a) the block-structured mesh
• (b) the trimmed cell mesh
• (c) the generic polyhedral mesh.

4
Allowed Mesh Types in Star-CCM
5
LES vs RANS

• Comparison of velocity magnitude distributions

• LES

• RANS

6
Initial Comparison of LES / RANS Results

• RANS using Star CD k-eps
• Close comparison of results ? RANS can predict
  cross-flow velocities in the wire-wrap case.
• Pressure-drop comparisons underway.

7
Allowed Mesh Types in Star-CCM
8
19-pin simulations

• Relative transverse velocity magnitude

9
Comparison of 3 Reynolds Numbers
Relative transverse velocity magnitude
10
37-pin Simulations

• Relative transverse velocity magnitude

11
Predicted Dimensionless Pressure Loss Coefficient
from RANS Simulations vs. Correlations

• The dimensionless pressure loss coefficient is
  the pressure drop normalized by the dynamic head,
  so that Cpf (L/D).
• The Cheng Todreas correlation assumes that
  there are three fundamental sub-channel types
  interior, edge, and corner. Each of the three
  types of sub-channel frictional losses is
  calculated separately. The bundle friction factor
  is then averaged.
• The Rehme correlation is a simpler single
  equation formulation based on representative
  geometric parameters.

Number of Pins Cheng Todreas Correlation Rehme Correlation RANS Simulation Prediction
7 1.116 14 1.179 5 2.282
19 1.088 14 1.041 5 1.199
37 1.075 14 0.943 5 1.059
Small 7-pin assemblies are not within the range
of applicability of the correlations
12
217-pin heated assembly

• Defined a typical SFR assembly for initial
  comparisons between CFD code predictions and
  SASSYS subchannel model predictions
• 217 pins
• 8 mm pin outer diameter
• Pitch-to-diameter ratio of 1.135
• 1.03 mm diameter wire wrap
• 20 cm wire wrap lead length
• Assumed power distribution
• Uniform radial power
• Cosine axial distribution
• Simplified boundary conditions
• Uniform inlet velocity
• 5.8 m/s
• Constant pressure outlet
• Specified wall heat flux on pin surfaces
• Adiabatic can surface

13
Axially coarsened polyhedral meshes
Nominal Cell Aspect Ratio Number of Cells Predicted Pressure Loss (kPa)
116 7507243 76.01
18 9769722 77.00
14 14955523 82.01
Changes in pressure loss are mostly form losses,
coarsening of surface representation may be
introducing artificial surface roughness.
14
Simplifed wire wrap representation

• Minimal effect on flow field
• Slight increase in size of low velocity region on
  leeward side of wire

Transverse velocity magnitude (inlet velocity
5.8 m/s)
15
217 pin SFR Assembly simulationPolyhedral mesh
with simplified wire wrap representation

• Coolant Temperature (K)

• Pin Surface Temperature (K)

16
217 pin SFR Assembly simulationPolyhedral mesh
with simplified wire wrap representation
17
MAX A Thermal Mixing Experiment

• LDRD funded project
• Define validation requirements for high-fidelity
  methods
• Evaluate/Demonstrate advanced measurement
  capabilities
• Develop methods for analysis and correlation of
  large 4-D data sets
• 10-100 TeraBytes per test
• Develop methods and visualization capabilities
  for validation comparisons of detailed 4-D data
• Provide data set for initial validation of
  individual advanced methods and integrated
  multi-scale code systems

18
RANS Simulations of MAX Thermal Mixing Experiments

• The computational model of the nominal geometry
  was developed from a simplified CAD
  representation
• The surfaces defined by the CAD model are
  triangulated and used as the basis for a
  tetrahedral mesh.
• The tetrahedral are then collapsed into generic
  polyhedra to form the mesh shown.
• Boundary Conditions
• uniform velocity, constant temperature inlets
• constant pressure outlet
• no-slip adiabatic walls
• Short turn-around time allows RANS to be applied
  to configuration studies

19
RANS Simulations of MAX Thermal Mixing Experiment

• Steady state simulations
• default segregated flow solver
• SIMPLE algorithm
• Rhie- Chow interpolation for pressure- velocity
  coupling
• algebraic multi-grid preconditioning.
• 1st-order upwind differencing scheme
• secondary gradient terms in the diffusion step
  were retained
• Realizable k-epsilon turbulence model with a
  two-layer all y wall treatment (Norris
  Reynolds)
• RANS simulations predict the development of a
  very sharp interface between the two jets
• Complex mixing pattern in the surrounding fluid
  after the jets have impinged on the upper surface.

20
Extracted Data

• To facilitate quantitative comparisons of
  characteristic data from simulations of different
  design options, data was extracted along lines at
  the mid-plane of the fish tank

21
Extracted Data - Nominal Configuration
22
Mesh resolution studies
23
MAX Experiment for TH Validation

• Initial RANS / LES Comparisons
• Average velocity distributions on two centerline
  cuts
• RANS 2 million gridpoints, steady state
• LES 23 million gridpoints, unsteady (still in
  initial transient)
• TKE comparisons similar

A
B
Pointer, Lomperski, Fischer., Validation of CFD
Methods for Advanced SFR Design Upper Plenum
Thermal Striping and Stratification,
ICONE17-75740, 2009
24
Other Leveraged Efforts

• Very High Temperature Reactor Integrated
  Multi-Physics Simulation
• 2-Phase Boiling Model
• Coarse Finite Element Design Simulations

25
VHTR Single Fuel Block Column

• Completed coupled CFD/Neutronics simluations of a
  single column of prismatic VHTR fuel blocks
• DeCART 2D/1D MOC
• STAR_CD Steady RANS
• CFD model uses 8.8 million computational cells
• Polyhedral elements allow conformal meshing of
  solid and fluid components
• Improved numerical performance for conjugate heat
  transfer
• Includes upper and lower plenum volumes
• Flow splits between channels are simulated rather
  than specified

26
DeCART/STAR-CD Mesh Mapping

• Initial mapping utility uses a simple approach in
  which DeCART zones are associated with all
  STAR-CD cells whose centroid falls within that
  zones.
• Global conservation is enforced within any single
  material across the entire domain

27
Coupled Full Block Model Results
T
q

• Power distribution from DeCART, reflects
  temperature feedback from CFD
• Temperature feedback exaggerated, due to greatly
  increased F/M ratio for single block
• After CFD initialization, coupled simulation
  required 4.2 hours on 32 cores for 9 data
  exchanges

28
2-Phase Boiling CFD

• Generation 1
• Bubbly flow topology used in all cells.
• Spherical vapour bubbles surrounded by continuous
  liquid phase.
• Used for previous BFBT calculations
• Generation 2
• Bubbly flow, Mist flow and Transition topologies
• Bubbly flow topology Spherical bubbles
  surrounded by liquid
• Mist flow topology Spherical droplets surrounded
  by vapor
• Preliminary liquid film models
• Used for current BFBT calculations
• Generation 3
• Bubbly flow, Mist Flow, Transition, and Sharp
  Interface topologies
• Uses 2-D Topology Map based on a and a gradients
• Inter-phase surface transport planned
• Developed jointly by Adapco, Argonne, Sarov Labs,
  VNIIEF

29
BFBT BWR Benchmark Simulation Results Exit
Quality 25
Microscopic Distribution (CT Scan Data)
Predicted void fraction
Percent Error
Channel Averages (Calculated from CT Scan Data)
Predicted void fraction
Percent Error
30
Coarse Finite Element Design Simulations

• Solve Convection-Diffusion
• Use imposed flow field based on RANS or LES
  simulations
• Relatively fast running
• A few minutes on a few hundred nodes
• High-order numerical accuracy
• High-order spatial accuracy

31
Duplicate spacer grid design experiment
methodology
32
Off-Set Injection Point
Effect of Nominal Spacer
Effect of Spacer With 0.5 H/D
No Spacers
Thermal Diffusion x 2
Thermal Diffusion x 4
33
CONCLUSIONS RANS (and other)

• The applicability of commercial RANS-based CFD
  tools to hydrodynamic analysis of wire-wrapped
  sodium-cooled fast reactor fuel assemblies has
  been demonstrated
• RANS-based CFD predictions compare well with LES
  predictions
• Sensitivity of predictions to Reynolds number is
  low
• Re9000 is probably sufficient for hydrodynamic
  studies
• Broad axial and radial mesh density sensitivity
  studies have been completed for 7-, 37-, and
  217-pin models
• Sensitivity of bulk predictions to mesh structure
  is low, but mesh structure impacts resolution of
  jet structures and recirculation regions on
  leeward side of pins.
• Completed initial benchmarking study comparing
  predicted pressure drop versus lumped parameter
  correlations and experimental data
• Completed study comparing predicted temperature
  distributions versus conventional sub-channel
  models based on lumped parameter correlations
• Demonstrated initial course mesh conduction
  diffusion solver

34
Current/Future Work RANS (and other)

• Comparisons with legacy experimental data in
  progress
• Turbulent diffusion in HEDL experiments
• 217 pin simulations with conjugate heat transfer
  in fuel, cladding and sodium bond gap
• Simulations of alternate spacer options
• Engineered can walls
• Reduced diameter edge row wire wrap
• Implementing thermophysical property functions
  for liquid sodium, clad and fuel.
• Beginning to extend simulations to ex-core
  regions
• Initiate development of standalone finite element
  subchannel simulation tool and finite element
  network flow solver

35
Questions?
36
Extra Slides
37
Central Injection Point
Effect of Spacer With 0.5 H/D
Effect of Nominal Spacer
No Spacers
Thermal Diffusion x 2
Thermal Diffusion x 4
38
Side Channel Injection Point
Effect of Spacer With 0.5 H/D
Effect of Nominal Spacer
No Spacers
Thermal Diffusion x 2
Thermal Diffusion x 4