1. ATTILA* Deterministic Code (Status and plans for validation)
Mahmoud Z Youssef
UCLA
Presented at ITER-TBM Meeting, May 10-11, 206
* Some of the viewgraphs are taken from Failla and Loughlin presentations (Frascati Meeting, Dec. 2005)

2. Topics
Deterministic versus Monte Carlo methods in nuclear analysis: Pros and Cons
ATTILA code:
Featues
Examples of its application
Status and plans for validating ATTILA for adoption in ITER nuclear analysis such as
Design of ITER-TBM ports
Design of ITER diagnostics port plugs

3. Features of Deterministic and Monte Carlo codes
Deterministic codes (e.g. DANTSYS,DOORS):
In solving Boltzman neutron balance equation neutron/g energy and angular direction are discretized (Muligroup, Sn). Cross-section are approximated with series of Legendre polynomials (Pn) and averaged over energy bins.
structured meshes are used to approximate complex 3D geometries (no mixing between different coordinate systems, e.g. rectangular, cylindrical).
n/g fluxes and associated reaction rates (tritium production, damage, etc.) are calculated everywhere in the system.
Monte Carlo codes (e.g. MCNP):
It is a stochastic process. Millions of source particles are followed in a random processes to estimate the required fluxes and associated responses at pre-selected locations (tallies). 3D complex geometries are described by combination of surfaces intersections to form bodies (zones). Point-wise nuclear data are used.

4. Advantages and Disadvantages
Deterministic codes
Advantages:
Fluxes and responses are calculated everywhere. No need to redo separate runs if additional responses are needed.
Shorter time to run a case compared to Monte Carlo methods.
Disadvantages:
Large disc space is required to store angular flux
Ray effect due to angular discretization
Cross section should be shielded particularly in resonance regions
Monte Carlo codes
Complex geometry can be modeled accurately. However, extensive effort is needed to generate the appropriate “geometry cards”. This is why CAD-based versions are in progress.
Fluxes and responses are calculated at pre-selected locations
Visualization of the responses in the whole system is not possible.

5. What is Attila?
A finite element Sn neutron, gamma and charged particle transport code using 3D unstructured grids (tetrahedral meshes)
Geometry input from CAD (Solid Works, ProE)
Supplied by Transpire Inc.

6. Version 5.0 – New Features Released November 15, 2005
Last Collided Flux Option for Post Processing
High angular resolution achievable, used for:
Resolving streaming paths
Evaluation of solution far from scattering media
Integrated Depletion Module
Functionality similar to ORIGEN
Automated Weight Windows Generation
Writes mc_wwinp file
User control over grid resolution
Post graphics and visualization in the entire system
Failla pres

7. 2006 – New Functionality Integrated Activation Capability
Extension of current depletion module to include decay source terms
Group-wise Adaptive SN Order
For ITER, can run 14 MeV source bin at a high Sn order to transport the primary flux
Distributed Memory Parallel
Linear scaling achieved on test version up to 256 processors
A primary motivation is to distribute memory resources
Failla pres

8. Benchmark At the exit to the labyrinth
To the side of the roof penetration.
Opposite the horizontal port in the mid-plane
50cm above the floor opposite the horizontal port
In the middle of the vessel wall
Loughlin pres.

9. At exit to labyrinth
Loughlin pres.

10. Model of ITER
Simplification was needed to reduce the number of bodies and the time taken to read in the geometry
Loughlin pres.

11. Simplified Model of ITER (with unstructured Meshes)
Approx 300,000 elements
For detailed ITER calculation s, estimate about 600,000 elements
Failla pres

12. Another Simplified model with a Diagnostic Port
Simplified Source location shown in white
Failla pres

13. Visualization of neutron flux
Iso-surface contour of the total flux
Failla pres

14. Visualization of neutron flux (2)
Iso-surface contour of the total flux
Failla pres

15. Analysis Geometry read into Attila Attila mesh: 424918 cells
Loughlin pres.

16. Results
Loughlin pres.

17. Results, Cont.
Failla pres

18. Results, cont.
Failla pres

19. ATTILA for Detailed ITER Calculations
Estimate about 600,000 elements needed for detailed calculations of a 40° section
- Would equate to approx 72 CPU hours
- Single collision high SN order for primary flux transport
- S18 scattering calculation
Results at any point can be extracted using the third order finite element spatial representation
Streaming in gaps can be done as a post processing step
Response functions calculated as a post processing step everywhere in the system.
Energy dependent flux in selected parts can be output for activation calculations
Failla pres

20. Two complimentary methods
Attila and MCNP
Two complimentary methods
ATTILA
- Scoping studies and design calculations
- Activation
- Global solution field data (mapping)
MCNP
Verification
Through the combination of CAD to MCNP translation and Attila adjoint based weight windows generation, it is now possible to perform two calculations (Attila and MCNP) faster than MCNP alone
Failla pres

21. Relative to Monte Carlo Attila CPU time << Monte Carlo
Computational Requirements
Relative to Monte Carlo
Attila CPU time << Monte Carlo
Much more data available from a single calculation
Very efficient for problems with:
- Large attenuations
- Activation
- Local field values desired (mapping)
Attila Memory consumption >> Monte Carlo
- Primary driver for parallelization
- Recommend 4 GB or more for ITER
-Disk storage requirements can be substantial
- Centralized computing system highly desirable
Failla pres

22. Status and Chronological Events for ATTILA Benchmarking (1)
October, 05: Attila was introduced to the fusion community in Garching by Gregory Failla (CEO of Transpire, Inc. )
December, 05: A meeting in Frascati, Italy, was held to discuss CAD-based codes. Fallia and Loughlin (UKAEA) showed comparison between MCNP and Attila for several cases. Franco Federici (ITER-IT) emphasized that for Attila to be adopted for ITER-related nuclear analysis, it must be benchmarked for quality QA purposes.
March, 06: Dave Johnson (US Diagnostic leader), in a meeting with M. Sawan and M. Youssef, emphasized the usefulness of Attila in the design of diagnostics port plugs. Hot spots can be easily identified from Attila visualization output plots.

23. Status and Chronological Events for ATTILA Benchmarking (2)
April 5, 06:
A teleconference was arranged by UCLA in which neutronics experts from both EU and US have participated. Failla, from Transpire Inc., gave an overview. Capabilities and limitations of Attila were discussed. It was decided to:
Benchmarking Attila for 3 integral experiments (tungsten, streaming, and bulk shielding) provided by Paola Batistoni (Frascati). Comparison will be made to MCNP results.
It was agreed that UCLA will undertake this benchmarking task

24. Two benchmarks experiments
One on the 3 benchmarks; Streaming Experiment
Two benchmarks experiments
Streaming Experiment
Bulk Shielding Experiment

25. Status and Chronological Events for ATTILA Benchmarking (3)
April 7, 06:
Logistics were discussed in a teleconference (Nelson, Johnson, Youssef, Sawan): They are:
- Four (4) neutronics and CAD persons from the US will be trained in WA on how to use Attila (may 31-June 2).
- Cost estimates for Attila benchmarking and training.
A meeting will was held in Madison, Wi, (July 24-26,06) to discuss:
- Progress in CAD-based Monte Carlo code development
- Attila benchmarking
IF passed QA, Attila can be very powerful tool in ITER-TBM nuclear design