WORKSHOP:ADVANTG: Automated

Variance Reductionfor MCNP

13:00–17:00August 26, 2018

Drury Plaza Hotel Santa Fe, NM, USA

For questions, contact: Scott Mosher, ORNLmoshersw@ornl.gov

Held in association with the20th Topical Meeting of the Radiation

Protection and Shielding Division

http://rpsd2018.ans.org/

ADVANTG is a software package that generates space- and energy-dependent weight window bounds and biased source probabilities that accelerate continuous-energy Monte Carlo simulations of fixed-course neutron and photos transport problems using MCNP. A key component of ADVANTG is the Denovo parallel discrete ordinates (SN) solver, which is used to generate the importance maps from which the variance reduction parameters are derived.

The ADVANTG team has developed and implemented two well-known variance methods, which are now in wide-spread use:• The Consistent Adjoint-Driven Importance Sampling (CADIS)

method for accelerating the convergence of individual tally estimates

• The Forward-Weighted CADIS (FW-CADIS) method for optimizing the simulation to obtain nearly uniform convergence of multiple tallies and mesh tallies

This workshop will highlight the capabilities of ADVANTG and cover the setup and execution of three tutorial problems: • a basic shielding problem,• a radiation detection problem, and • a deep penetration global variance reduction problem.

The open-source VisIt 3-D visualization software will be used as an integral part of these tutorials. The session will also include a summary of our experience applying ADVANTG in high-fidelity radiation shielding and material damage analyses of the High Flux Isotope Reactor, Spallation Neutron Source, and ITER.

Neutron fluxes throughout ITER facility

Workshop Schedule

13:00-14:15 CADIS and FW-CADIS methods and ADVANTG implementation

The theoretical framework that underlies the variance reduction capabilities of ADVANTG and the algorithms that were developed to implement the theory will be described. ADVANTG implements the Consistent Adjoint Driven Importance Sampling (CADIS) method to accelerate individual tallies and the Forward-Weighted CADIS (FW-CADIS) method to obtain nearly uniform statistical precision across multiple tallies and large mesh tallies. These methods provide a prescription for generating space- and energy-dependent weight windows and biased source probabilities based on approximate solutions to the adjoint transport equation. The implementation of the methods will be discussed to provide a solid understanding of the inputs, outputs, and overall operation of ADVANTG.

14:15–15:15 ADVANTG demonstration and VisIt visualization

The use of ADVANTG to accelerate a single tally using the CADIS method and to obtain nearly uniform convergence of a mesh tally using the FW-CADIS method will be demonstrated. This presentation will include the visualization of the discretized geometry and discrete ordinates solutions using the open-source VisIt software (https://wci.llnl.gov/codes/visit/). MCNP results estimated with and without the ADVANTG-generated variance reduction parameters will be compared and discussed.

15:15–15:45 Recent ADVANTG applications at ORNL

A summary of recent radiation shielding applications performed at ORNL using ADVANTG and MCNP will be presented. Applications will include analyses of neutron and gamma fluence and material damage to reactor pressure vessels and biological dose-rate distributions in fusion reactor facilities. The primary purpose of this session is to provide examples of applying ADVANTG in large-scale, real-world transport applications and the quality of results that can be obtained.

15:45–16:30 Beyond the basics: Advanced tips and suggestions

Detailed tips and suggestions for generating effective variance reduction parameters using ADVANTG will be presented. Common pathologies of underperforming variance reduction parameters will be illustrated using simple example problems along with the steps needed to improve the parameters. Fundamental to this discussion is an understanding of the statistical tests implemented in MCNP that indicate tally convergence behavior. Suggestions for reviewing and acting upon this information will be discussed.

16:30–17:00 Soon-to-be-released capabilities

This final session will provide a preview of new features and improvements that will be released in the next version of ADVANTG and describe two new related capabilities. The parallel Radiant 3-D image generator has been developed to produce accurate, high-quality ray-traced images of MCNP geometries at arbitrary resolution. The ORNL Transformative Neutronics (ORNL-TN) upgrade to MCNP5 was developed to implement a shared-memory mesh tally capability that significantly increases the upper limit of mesh tally resolution in multithreaded simulations. ORNL-TN also provides fast geometry initialization of highly detailed MCNP models, such as those converted and compiled from multiple CAD models.