A Gaussian-Process Surrogate for Control Rod Worth with Calibrated Uncertainty: Application to VERA Core Physics Benchmark Problem 4
Abstract
Differential and integral control rod worth are usually obtained by differencing eigenvalue calculations at discrete rod positions, which resolves the derivative poorly and gives it no uncertainty. We build a Gaussian-process (GP) surrogate for reactivity as a function of rod position on VERA Core Physics Benchmark Problem 4, a 3×3 assembly cluster with a hybrid AIC/B4C rod cluster control assembly, modelled in OpenMC with ENDF/B-VIII.0. A parametric sigmoidal mean carries the dynamic range, a Matern-5/2 GP carries the residual, and the reported Monte Carlo uncertainty of each run enters as known observation noise. Because differentiation is linear, the differential rod worth and its variance follow in closed form from the kernel derivatives. Trained on 43 runs at three particle counts and tested on 11 independent runs, the surrogate meets a fixed 10 pcm accuracy target above the test-run noise, and its intervals are calibrated: the standardized leave-one-out residuals over 129 independent runs have variance 1.07 with 95% interval [0.85, 1.39]. The total worth agrees with the direct OpenMC end-point difference within combined uncertainty, and the differential worth peak is located to within one step. Three mean functions of very different fit quality give the same differential worth to within 0.1 pcm/step, while the width of its band depends on the kernel smoothness by a factor of four. The number of runs required and the remaining approximations are reported, and the extension to rod position and soluble boron is outlined.
Related
- A physics-informed uncertainty-aware surrogate for p-n junction diodes
- ONE CYLinder: A Benchmark for Graph-Based Surrogate Modeling of Unsteady Bluff-Body Flows
- Reconstruction of Unobserved Core Electron Density in Tokamak Plasmas with a Neural Operator Transport Surrogate and Uncertainty Quantification
- A differentiable PCA-GPR surrogate framework for full-transient Bayesian inverse uncertainty quantification of black-box nuclear thermal-hydraulic systems
- ELMO: An Uncertainty-Aware Simulation-to-Surrogate Workflow for Fast Pedestal Linear-Stability Prediction
BibTeX
@misc{gohlke2026gaussianprocess,
title = {{A Gaussian-Process Surrogate for Control Rod Worth with Calibrated Uncertainty: Application to VERA Core Physics Benchmark Problem 4}},
author = {Sean Gohlke and Jilang Miao},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.40389},
eprint = {2610.40389},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.40389},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - A Gaussian-Process Surrogate for Control Rod Worth with Calibrated Uncertainty: Application to VERA Core Physics Benchmark Problem 4
AU - Sean Gohlke
AU - Jilang Miao
PY - 2026
DA - 2026/10/29
PB - arXiv
DO - 10.48550/arxiv.2610.40389
UR - https://arxiv.org/abs/2610.40389
AB - Differential and integral control rod worth are usually obtained by differencing eigenvalue calculations at discrete rod positions, which resolves the derivative poorly and gives it no uncertainty. We build a Gaussian-process (GP) surrogate for reactivity as a function of rod position on VERA Core Physics Benchmark Problem 4, a 3×3 assembly cluster with a hybrid AIC/B_4C rod cluster control assembly, modelled in OpenMC with ENDF/B-VIII.0. A parametric sigmoidal mean carries the dynamic range, a Matern-5/2 GP carries the residual, and the reported Monte Carlo uncertainty of each run enters as known observation noise. Because differentiation is linear, the differential rod worth and its variance follow in closed form from the kernel derivatives. Trained on 43 runs at three particle counts and tested on 11 independent runs, the surrogate meets a fixed 10 pcm accuracy target above the test-run noise, and its intervals are calibrated: the standardized leave-one-out residuals over 129 independent runs have variance 1.07 with 95% interval [0.85, 1.39]. The total worth agrees with the direct OpenMC end-point difference within combined uncertainty, and the differential worth peak is located to within one step. Three mean functions of very different fit quality give the same differential worth to within 0.1 pcm/step, while the width of its band depends on the kernel smoothness by a factor of four. The number of runs required and the remaining approximations are reported, and the extension to rod position and soluble boron is outlined.
ER -