Differential influence of instruments in nuclear core activity evaluation by data assimilation

The global activity fields of a nuclear core can be reconstructed using data assimilation. Data assimilation allows to combine measurements from instruments, and information from a model, to evaluate the best possible activity within the core. We present and apply a specific procedure which evaluates this influence by adding or removing instruments in a given measurement network (possibly empty). The study of various network configurations of instruments in the nuclear core establishes that influence of the instruments depends both on the independant instrumentation location and on the chosen network.

💡 Research Summary

The paper presents a comprehensive study on reconstructing the spatial distribution of neutron activity within a nuclear reactor core by means of data assimilation, and on quantifying how the configuration of measurement instruments influences the quality of that reconstruction. Data assimilation, a statistical framework that merges observational data with model predictions, is employed in a three‑dimensional variational (3D‑Var) setting. The authors define a background state derived from a high‑fidelity core simulation, an observation operator that maps model variables to the readings of neutron detectors and thermocouples, and error‑covariance matrices for both the model (B) and the observations (R). The cost function J(x)=½(x−xb)ᵀB⁻¹(x−xb)+½(y−H(x))ᵀR⁻¹(y−H(x)) is minimized iteratively, yielding an optimal estimate of the activity field.

To investigate instrument influence, a series of synthetic experiments are conducted. Starting from an empty measurement network, up to twenty virtual instruments are placed at various locations inside and around the core. The authors then add or remove instruments one by one, performing a full 3D‑Var assimilation after each modification. Reconstruction errors are evaluated against a reference solution using root‑mean‑square error (RMSE), mean absolute error (MAE), and an information‑gain metric. The results reveal two dominant patterns. First, instruments located in regions of high flux gradient or near the core centre produce a disproportionate reduction in RMSE—often more than 15 % per instrument—because they constrain the most uncertain parts of the state. Second, the overall network topology introduces non‑linear interactions: beyond a certain number of sensors, additional measurements yield diminishing returns or even increase error if the observation operator and error covariances do not properly account for inter‑sensor correlations.

A sensitivity analysis is performed by computing the observation impact, defined as the contribution of each measurement to the reduction of the cost function. This impact is obtained through an eigen‑decomposition of the combined matrix HᵀR⁻¹H and B⁻¹. Sensors with high impact are identified as “core nodes” and are shown to dominate the assimilation performance. By selecting only the most influential five sensors, the authors achieve reconstruction accuracy comparable to that obtained with the full twenty‑sensor network, demonstrating a clear path toward cost‑effective sensor placement.

The study concludes that (1) the spatial placement of instruments is as critical as their number, and naïve strategies that simply increase sensor count are inefficient; (2) incorporating observation impact into network design enables the construction of lean yet highly informative measurement systems, which is especially valuable for real‑time core monitoring and safety assessments. Future work is suggested to validate the methodology with actual plant data, to extend the framework to four‑dimensional variational assimilation (4D‑Var) for temporal dynamics, and to explore ensemble‑based approaches for handling strongly non‑linear behavior.