From Models To Experiments: Shallow Recurrent Decoder Networks on the DYNASTY Experimental Facility

TL;DR

This work tackles high-dimensional state estimation for nuclear-thermal-hydraulic systems using a data-driven surrogate, SHRED, that blends LSTM-based temporal modeling, a shallow decoder, and SVD compression to infer full-state fields from sparse sensor data. Applied to the DYNASTY natural-circulation facility, SHRED is trained on RELAP5 high-fidelity data and validated against real temperature measurements, showing strong parametric interpolation and forecasting capabilities with fast training times. The results demonstrate that SHRED can locally correct imperfect background models using online sensor inputs and provide accurate state estimates in regions covered by sensors, offering a practical, lightweight tool for monitoring and control in complex engineering systems. Open-source code supports reproducibility and adaptation to other parametric or multi-physics scenarios.

Abstract

The Shallow Recurrent Decoder networks are a novel paradigm recently introduced for state estimation, combining sparse observations with high-dimensional model data. This architecture features important advantages compared to standard data-driven methods including: the ability to use only three sensors (even randomly selected) for reconstructing the entire dynamics of a physical system; the ability to train on compressed data spanned by a reduced basis; the ability to measure a single field variable (easy to measure) and reconstruct coupled spatio-temporal fields that are not observable and minimal hyper-parameter tuning. This approach has been verified on different test cases within different fields including nuclear reactors, even though an application to a real experimental facility, adopting the employment of in-situ observed quantities, is missing. This work aims to fill this gap by applying the Shallow Recurrent Decoder architecture to the DYNASTY facility, built at Politecnico di Milano, which studies the natural circulation established by internally heated fluids for Generation IV applications, especially in the case of Circulating Fuel reactors. The RELAP5 code is used to generate the high-fidelity data, and temperature measurements extracted by the facility are used as input for the state estimation. The results of this work will provide a validation of the Shallow Recurrent Decoder architecture to engineering systems, showing the capabilities of this approach to provide and accurate state estimation.

Figures (7)

• Figure 1: SHRED architecture applied to the DYNASTY facility. Three thermocouples are used to measure the temperature in the fluid $T$. The sensors time series are used to construct a latent temporal sequence model which is mapped to the compressive representations of all spatio-temporal field variables. The compressive representations can then be mapped to the original state space by the Singular Value Decomposition (SVD).
• Figure 2: DYNASTY natural circulation loop cammi2019dynamics: on the left, a picture of the real facility is provided, whereas on the right the scheme of the system is reported with the main components of the facility.
• Figure 3: R5 nodalization of the DYNASTY experimental facility. The red zone corresponds to the heated section for the VHHC-GV1 case, whereas the blue zone corresponds to the finned cooler.
• Figure 4: Train (75%) -validation (12.5%) - test (12.5%) split of the parametric space, composed by the power provided to each control volume $P$ and the heat transfer coefficient $h$ at the cooler. The experimental configuration at $\boldsymbol{\mu}^{\text{exp}} = \left[35.5\,\text{W}, \; 65.0 \,\frac{\text{W}}{\text{m}^2\text{ K}}\right]$ for the GV1 experiment is also displayed.
• Figure 5: Verification of SHRED using synthetic data only for parametric scenarios at $\boldsymbol{\mu}^{\text{exp}} = \left[35.5\,\text{W}, \; 65.0 \,\frac{\text{W}}{\text{m}^2\text{ K}}\right]$: contour plots of the FOM and SHRED are shown on the left, whereas the dynamical evolution of the temperature at the experimental locations is shown on the right.