Neutronics for system codes

The 25th European Fusion Programme Workshop

Dubrovnik, Croatia, 27-29 November 2017

Neutronics for system codes.

by Jonathan Shimwell

To perform automated parametric multiphysics analysis of breeder blanket designs with an aim of optimising the design.

Objective

To perform automated parametric multiphysics analysis of breeder blanket designs with an aim of optimising the design.

Neutronics for system codes.

by Jonathan Shimwell

Selection of design parameters

Parametric CAD construction

Neutronics simulation for TBR and EM

Conversion to unstructured mesh

Neutronics simulation for volumetric heating

Conversion to engineering mesh

Simulations to find stress and temperature

Evaluate design

Conversion to neutronics model

Interpolate performance & design sensitivity

Model generation

PythonOCC

Salome

Parametric

HCPB model

FreeCAD

Parametric

HCLL model

Neutronics model

MCNP

constructive

solid geometry

MCNP

unstructured

mesh geometry

Serpent

STL geometry

Serpent

unstructured

mesh geometry

Engineering

model

Parafem

model

Code Astra

model

Trelis

Hex mesh

Abaqus

Trelis

Hex mesh

OpenFoam

STEP to STL conversion

Geometry conversion

MCam / McCad

CAD to CSG

converters

Meshing

Trelis

Tet mesh

Ensight Gold

GMesh

Salome

MED

Avoid the use of constructive

solid geometry (CSG)

CAD model

Convert to CSG model

Assign materials

Parameterise CSG model

Neutronics simulation

CAD model

Convert to STL model

Assign materials

Parameterise CAD model

Neutronics simulation

No spline support. Void Generation

Lack of tools

Cell based

Rapid and controllable

Mature software

File based

Optimal parameters

CAD

model

Traditional CSG work flow

Alternative STL work flow

Blanket geometry construction

Blanket envelop input

Feature recognition

Output geometry

Front face

Edge to fillet

Top face

Variable detail blanket geometry

Surface deviation = 0.01mm

Surface deviation = 1mm

Homogenised first wall

Detailed first wall

Blanket geometry construction

Customizable cooling pipe layouts for the internal cooling plates

Blanket geometry construction

Blanket envelopes from the entire reactor can be parameterised

HCLL blanket geometry parameter options

Armour thickness
First wall thickness
First wall cooling channels (height, width, pitch)
End caps thickness
Lithium lead poloidal distance
Cooling plates poloidal distance
Cooling plate coolant channels (height width)
Toroidal distance of rear components (plate, lithium lead, coolant, divider and cap)

HCPB blanket geometry parameter options

Lithium ceramic poloidal distance

Neutron multiplier poloidal distance

Choice of breeder material Li4SiO4, Li2SiO3, Li2ZrO3, Li2TiO3

Choice of neutron multiplier Be, Be12Ti, Ba5Pb3, Nd5Pb4, Zr5Pb3, Zr5Pb4

Computing techniques

Cloud computing and containerization of code was used for this project.

Containerization allows for codes to be combined with their dependencies packaged into a container and deployed on compatible infrastructure.

Serpent II

Nuclear data

Docker

Computing techniques

Docker

Singularity

Swarm

Orchestration

Compute

Containerize

Storage

Day of the week

Cluster utilisation

100%

Your local computing cluster?

Mon

Tue

Wed

Thu

Fri

Sat

Sun

Example parameter study

Interpolation of results

Li4SiO4 with 60% 6Li and Be

Tritium production

Li4SiO4 at 60% 6Li enrichment with Be

Fitting the data using Gaussian Processes and training datasets

The Gaussian functions found fits the provided data points very well?

Fitting the data using Gaussian Processes and training datasets

Red points are used for fitting

Blue points are used for testing the fit

Volumetric heating simulations

Due to the use of CAD throughout the project the same model can be used to obtain TBR, volumetric heating, temperature and stress.

Procedure for obtaining volumetric heating

Neutron and photon heating simulated and results passed to engineering codes

Parametric CAD models created using Python scripts and automatically converted to neutronics models

Unstructured mesh applied to the geometry

Automated hex and conformal tet meshing

Neutronics mesh?

Engineering mesh

Do we need meshing algorithms designed for neutronics tallies?

Mesh element density should be concentrated in areas of high gradient.

Simulation time increases with number of mesh elements.

Meshing

Fully automated hex meshing of the geometry. Exporting to Abaques format for use with MCNP 6 unstructured mesh and Open Foam format for use in Serpent II.

Volumetric heating

Heating values obtained on unstructured mesh that conforms to material boundaries.