Addressing A Posteriori Performance Degradation in Neural Network Subgrid Stress Models

TL;DR

This paper explores methods to improve the real-world performance of neural network subgrid stress models in Large Eddy Simulations by combining data augmentation and input simplification, resulting in more robust and reliable models.

Contribution

It introduces a combined approach of training data augmentation and input complexity reduction to bridge the gap between a priori and a posteriori neural network performance in LES.

Findings

Training with multiple filters does not degrade a priori performance.

Models trained with augmented data are more robust across different LES codes.

Simplifying input features reduces discrepancies between a priori and a posteriori results.

Abstract

Neural network subgrid stress models often have a priori performance that is far better than the a posteriori performance, leading to neural network models that look very promising a priori completely failing in a posteriori Large Eddy Simulations (LES). This performance gap can be decreased by combining two different methods, training data augmentation and reducing input complexity to the neural network. Augmenting the training data with two different filters before training the neural networks has no performance degradation a priori as compared to a neural network trained with one filter. A posteriori, neural networks trained with two different filters are far more robust across two different LES codes with different numerical schemes. In addition, by ablating away the higher order terms input into the neural network, the a priori versus a posteriori performance changes become less apparent. When combined, neural networks that use both training data augmentation and a less complex set of inputs have a posteriori performance far more reflective of their a priori evaluation.