What question did this study set out to answer?

The central aim is to create a digital twin framework to predict temperature fields in nuclear thermal propulsion reactors efficiently.

March 26, 2026Nuclear Techniques0 citationsOpen Access

Temporal prediction of parameters in nuclear thermal propulsion reactors via latent space nonlinear operator learning

YZYiling ZENG JMJunyu MO WLWei Li

Key Points

The central aim is to create a digital twin framework to predict temperature fields in nuclear thermal propulsion reactors efficiently.
Constructed a lightweight 'encoding–operator learning–decoding' architecture.
Compressed temperature field data into a 100-dimensional latent space.
Used an improved Deep Operator Network (DeepONet) to model dynamic evolution in the latent space.
Applied a physics-constrained decoder for precise reconstruction.
Validated predictions using a nuclear thermal coupling dataset from OpenFOAM.
Average relative error in fuel region temperature predictions is below 1%.
Coolant temperature field error is below 0.5%.
Maximum relative error in 100-s predictions is under 0.5%.
Each predictive iteration executes in under 200 seconds, achieving over 500 times acceleration over traditional methods.
Single long-step predictions completed in 79.23 seconds, meeting real-time requirements.

Abstract

BackgroundIn nuclear thermal propulsion reactor engineering, the real-time prediction of multi-physics temperature fields under transient operating conditions has long been hindered by the inherent computational inefficiency and high dimensionality of coupled neutronics-thermal-hydraulic simulations. Traditional computational fluid dynamics (CFD) methods for required accurate, struggle to meet real-time decision-making requirements due to their prohibitive computational costs.PurposeThis study aims to develop a lightweight digital twin framework for achieving real-time, high-fidelity temperature field predictions in nuclear reactors, thereby enabling rapid safety assessment and control strategy optimization.MethodsFirst of all, a lightweight architecture of "encoding–operator learning–decoding" was constructed, and temperature field data of 230 000~1.8 million dimensions were compressed into a 100-dimensional latent space. Then an improved Deep Operator Network (DeepONet) was used to learn the dynamic evolution within this latent space, and a physics-constrained decoder was employed for high-precision reconstruction. Finally, a nuclear thermal coupling benchmark dataset generated by the computational fluid dynamics software OpenFOAM was applied to verifying prediction data.ResultsValidation results show that, in 40-s iterative predictions, the average relative error of the fuel region temperature field is below 1%, and the coolant temperature field error is below 0.5%; in 100-s long-step predictions, the maximum relative error remains under 0.5%. Each iteration of the predictive model takes less than 200 s, achieving over 500 times acceleration compared to traditional CFD methods; a single long-step prediction takes 79.23 s, which is shorter than the prediction interval, thus meeting real-time prediction requirements.ConclusionsThis study provides a lightweight digital twin solution for transient condition simulation and safety control of nuclear thermal propulsion reactors.

问 AI

Bookmark

View Full Paper