Node Assigned physics-informed neural networks for thermal-hydraulic system simulation: CVH/FL module

TL;DR

This work addresses the need for physics-consistent surrogate solvers in nuclear TH simulations by developing a Python emulator for MELCOR’s CVH/FL module and introducing a data-free PINN framework. A vanilla PINN struggled to learn transient CVH/FL dynamics, prompting the node-assigned PINN (NA-PINN) approach, where separate networks model each CV/FL quantity while a shared physics-based loss enforces coupling. NA-PINN significantly outperforms the original PINN across systems with 2, 3, and 6 CVs, achieving substantially lower MSE and demonstrating stable training where the single-network PINN failed. The results suggest a viable path toward physics-informed, multi-physics surrogate solvers that can operate without labeled data and scale with TH system complexity, with future work extending to additional MELCOR packages and real-time surrogate capabilities.

Abstract

Severe accidents (SAs) in nuclear power plants have been analyzed using thermal-hydraulic (TH) system codes such as MELCOR and MAAP. These codes efficiently simulate the progression of SAs, while they still have inherent limitations due to their inconsistent finite difference schemes. The use of empirical schemes incorporating both implicit and explicit formulations inherently induces unidirectional coupling in multi-physics analyses. The objective of this study is to develop a novel numerical method for TH system codes using physics-informed neural network (PINN). They have shown strength in solving multi-physics due to the innate feature of neural networks-automatic differentiation. We propose a node-assigned PINN (NA-PINN) that is suitable for the control volume approach-based system codes. NA-PINN addresses the issue of spatial governing equation variation by assigning an individual network to each nodalization of the system code, such that spatial information is excluded from both the input and output domains, and each subnetwork learns to approximate a purely temporal solution. In this phase, we evaluated the accuracy of the PINN methods for the hydrodynamic module. In the 6 water tank simulation, PINN and NA-PINN showed maximum absolute errors of 1.678 and 0.007, respectively. It should be noted that only NA-PINN demonstrated acceptable accuracy. To the best of the authors' knowledge, this is the first study to successfully implement a system code using PINN. Our future work involves extending NA-PINN to a multi-physics solver and developing it in a surrogate manner.

Figures (12)

• Figure 1: Nodalization for the scenario model
• Figure 2: Numerical scheme of calculation process utilized in MELCOR
• Figure 3: Velocity calculation process in MELCOR
• Figure 4: Velocity profiles in the FL computed by MELCOR (left) and emulator model (right)
• Figure 5: Water level profiles in the water tanks computed by MELCOR (left) and emulator model (right)