Comparison of CNN-based deep learning architectures for unsteady CFD acceleration on small datasets

TL;DR

This paper compares CNN-based deep learning models for accelerating unsteady CFD simulations with small datasets, highlighting ConvLSTM-UNet's superior performance and the challenges of error accumulation for long-term predictions.

Contribution

It provides a fair comparison of advanced CNN architectures for CFD acceleration, demonstrating their potential and limitations in small data scenarios within the RePIT framework.

Findings

ConvLSTM-UNet outperforms other models in accuracy and stability.

Error accumulation limits reliable predictions to about 10 timesteps.

Future work includes exploring graph neural networks and implicit neural representations.

Abstract

CFD acceleration for virtual nuclear reactors or digital twin technology is a primary goal in the nuclear industry. This study compares advanced convolutional neural network (CNN) architectures for accelerating unsteady computational fluid dynamics (CFD) simulations using small datasets based on a challenging natural convection flow dataset. The advanced architectures such as autoencoders, UNet, and ConvLSTM-UNet, were evaluated under identical conditions to determine their predictive accuracy and robustness in autoregressive time-series predictions. ConvLSTM-UNet consistently outperformed other models, particularly in difference value calculation, achieving lower maximum errors and stable residuals. However, error accumulation remains a challenge, limiting reliable predictions to approximately 10 timesteps. This highlights the need for enhanced strategies to improve long-term prediction stability. The novelty of this work lies in its fair comparison of state-of-the-art CNN models within the RePIT framework, demonstrating their potential for accelerating CFD simulations while identifying limitations under small data conditions. Future research will focus on exploring alternative models, such as graph neural networks and implicit neural representations. These efforts aim to develop a robust hybrid approach for long-term unsteady CFD acceleration, contributing to practical applications in virtual nuclear reactor.