Efficient neural models speed up radiation transport simulations in engineering
Efficient Neural Surrogates for Linear Radiation Transport on the Lattice and Hohlraum benchmarks
Artificial Intelligence
Summary
Simulating how radiation moves through materials is very important but can be very slow and expensive. The authors show that special neural network models can approximate these complex simulations much faster. They compare two advanced neural approaches and find that how the models are built and trained affects their accuracy differently depending on the problem. Their work helps guide choosing or designing neural surrogates tailored to specific radiation transport tasks. They also share their code and data to help others use and improve these faster simulation tools.
What this means in practice
- •For nuclear engineers: Speed up design optimization by replacing costly radiation transport solvers with trained neural surrogates tailored to specific simulation tasks.
- •For medical imaging software developers: Incorporate efficient neural approximations of radiation transport to accelerate image reconstruction processes that rely on complex simulations.
Authors
Carmelo Gonzales, Steffen Schotthöfer, Cory D. Hauck
Abstract
Linear radiation transport equations (RTEs) form the simulation foundations underpinning design and analysis tasks in nuclear engineering, inertial confinement fusion, medical imaging, and astrophysics, but resolving the high-dimensional phase space at engineering fidelity remains expensive enough that outer-loop workflows, such as design optimization, uncertainty quantification, and parameter sweeps, are routinely budget-bound on traditional solvers. Neural surrogates promise to relax this bottleneck by amortizing simulation cost across thousands of downstream queries, but the architectural choices and engineered inductive biases that make a surrogate accurate on one transport problem do not transfer straightforwardly across model families. We benchmark two parameter-matched neural surrogate architectures, the physics-attention Transolver and the multi-scale graph network Bi-Stride Multi-Scale MeshGraphNet (BSMS-MGN), as end-to-end approximations of the final-time particle concentration for the two-dimensional linear RTE on the canonical Lattice and Hohlraum benchmarks. An ablation across Fourier features and region-weighted training loss exposes strongly architecture-dependent inductive-bias preferences, indicating that design choices common to physics-informed surrogate workflows must be revisited per architecture rather than imported across model families, and that downstream utility depends on per-QoI sensitivity rather than a single field-level score. The model training recipe, training data, and evaluation pipeline are released alongside this paper to support reproduction, transfer to related transport problems, and evaluation as amortized forward-model components in larger outer-loop simulation workflows.