Data-driven three-dimensional super-resolution imaging of a turbulent jet flame using a generative adversarial network

Abstract

Three-dimensional (3D) computed tomography (CT) is becoming a well-established tool for turbulent combustion diagnostics. However, the 3D CT technique suffers from contradictory demands of spatial resolution and domain size. This work therefore reports a data-driven 3D super-resolution approach to enhance the spatial resolution by two times along each spatial direction. The approach, named 3D super-resolution generative adversarial network (3D-SR-GAN), builds a generator and a discriminator network to learn the topographic information and infer high-resolution 3D turbulent flame structure with a given low-resolution counterpart. This work uses numerically simulated 3D turbulent jet flame structures as training data to update model parameters of the GAN network. Extensive performance evaluations are then conducted to show the superiority of the proposed 3D-SR-GAN network, compared with other direct interpolation methods. The results show that a convincing super-resolution (SR) operation with the overall error of ∼ 4 % and the peak signal-to-noise ratio of 37 dB can be reached with an upscaling factor of 2, representing an eight times enhancement of the total voxel number. Moreover, the trained network can predict the SR structure of the jet flame with a different Reynolds number without retraining the network parameters.

Unsupervised learning based noise reduction algorithm in 2D Rayleigh images

“…To better highlight the performance superiority of the proposed model, several neural networks commonly used in the denoising field, such as DNGAN [8] , RESTNET [12] and DNCNN [13] , which all belong to supervised learning. Simultaneously, we introduce PSNR [14] and ER [7] to comprehensively evaluate the denoising results of several networks to make the evaluation results more convincing. As shown in Figure 5, the evaluation includes 20 times instants, and four networks were trained and tested with the same dataset.…”

“…With the rapid development of artificial intelligence, the algorithms based on deep learning has been gradually used in image denoising and reconstruction. Recently, Xu et al proposed a 3D super resolution generative adversarial network (3D-SRGAN) of a turbulent jet flame [7], which successfully applied deep learning technique to laser diagnostics. Cai et al proposed a denoising generative adversarial network (DNGAN) based on supervised learning [8] , which achieved denoising of two dimensional Rayleigh images.…”

Super-resolution analysis via machine learning: a survey for fluid flows

“…Moreover, the use of a CNN-based GAN for three-dimensional super-resolution analysis was examined by Xu et al [160] for computed tomography (CT) of turbulent jet combustor. With an example of turbulent atmospheric flow, Hassanaly et al [161] has comprehensively compared various models for super-resolution reconstruction, including a super-resolution GAN [162], stochastic estimation, a deconvolution GAN [163], and diversity-sensitive conditional GAN [164].…”

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

“…The goal is of the current work is to use image SR techniques to increase the resolution of the fluid phase fraction generated by a solver that uses the Volume Of Fluids (VOF) method of modelling two-phase incompressible turbulent fluid flow. We make use of the InterFoam solver from OpenFOAM [33], an open-source and highly performant CFD software library commonly used in both industry and research [34]. Upsampling the output approximates the results of increasing the density of the discretized mesh, and ideally will produce additional detail without incurring the cost of computation that comes with direct numerical solutions.…”