Using Physics-Informed Generative Adversarial Networks to Perform Super-Resolution for Multiphase Fluid Simulations

Abstract

Computational Fluid Dynamics (CFD) simulations are useful to the field of engineering design as they provide deep insights on product or system performance without the need to construct and test physical prototypes. However, they can be very computationally intensive to run. Machine learning methods have been shown to reconstruct high-resolution single-phase turbulent fluid flow simulations from low-resolution inputs. This offers a potential avenue towards alleviating computational cost in iterative engineering design applications. However, little work thus far has explored the application of machine learning image super-resolution methods to multiphase fluid flow (which is important for emerging fields such as marine hydrokinetic energy conversion). In this work, we apply a modified version of the Super-Resolution Generative Adversarial Network (SRGAN) model to a multiphase turbulent fluid flow problem, specifically to reconstruct fluid phase fraction at a higher resolution. Two models were created in this work, one which incorporates a physics-informed term in the loss function and one which does not, and the results are discussed and compared. We found that both models significantly outperform non-machine learning upsampling methods and can preserve a substantial amount of detail, showing the versatility of the SRGAN model for upsampling multi-phase fluid simulations. However, the difference in accuracy between the two models is minimal indicating that, in the context studied here, the additional complexity of a physics-informed approach may not be justified.

Surface Line Integral Convolution-Based Vortex Detection Using Computer Vision

“…The recent advances made in the őeld of machine learning have seen a multitude of applications for physics and simulation-based problems. Li et al [24] proposed to use a super-resolution generative adversarial network (SRGAN) to reconstruct ŕuid phase fractions in turbulent, multiphase ŕows. Ayli et al [25] used machine learning for active ŕow control applied to the ŕow around a circular cylinder.…”

Super-resolution and uncertainty estimation from sparse sensors of dynamical physical systems

“…Their introduction of a Downsampled Skip-Connection Multi-Scale (DSC/MS) model (Fukami et al, 2019), which features skip connections between the downsampling and upsampling blocks of the FCNN, showed superior results than CNNs in reconstructing turbulent flows. Supervised, self-supervised, and unsupervised methods have also been developed using CNNs (Maulik et al, 2019;Gao et al, 2021;Zayats et al, 2022) and GANs (Bode et al, 2019(Bode et al, , 2021Li and McComb, 2022) augmented with physics-informed (PI) loss functions.…”

Focus and Modality: Defining a Roadmap to Future Ai-Human Teaming in Design

“…There are different ways to obtain the data you need to train on a situation: from prior studies and experience, synthetically by creating a digital twin and generating design solutions, by transferring data from other problems that may not be direct matches but that have some overlapping properties, and data augmentation especially for geometries that can be rotated, translated and stretched. Emerging studies highlight the ways in which data limits machine learning potential, while other studies highlight ways to minimize data usage by encoding expert knowledge (Maier, Soria Zurita, et al, 2022) or physics (Li and McComb, 2022;Pierce et al, 2021).…”