Physics-informed neural networks (PINNs) are successful machine-learning methods for the solution and identification of partial differential equations (PDEs). We employ PINNs solving Reynolds-averaged Navier$\unicode{x2013}$Stokes (RANS) incompressible turbulent flows without any specific model or assumption turbulence, by taking only data on domain boundaries. first show applicability laminar Falkner$\unicode{x2013}$Skan boundary layer. then apply simulation four turbulent-flow cases, i.e.,...

The long short-term memory (LSTM) neural network is used to predict the temporal evolution of a low-order representation near-wall turbulence. This leads excellent predictions turbulence statistics and system dynamics, characterized by Poincar\'e maps Lyapunov exponents.

Two models based on convolutional neural networks are trained to predict the two-dimensional instantaneous velocity-fluctuation fields at different wall-normal locations in a turbulent open-channel flow, using wall-shear-stress components and wall pressure as inputs. The first model is fully network (FCN) which directly predicts fluctuations, while second one reconstructs flow linear combination of orthonormal basis functions, obtained through proper orthogonal decomposition (POD), hence...

This work evaluates the applicability of super-resolution generative adversarial networks (SRGANs) as a methodology for reconstruction turbulent-flow quantities from coarse wall measurements. The method is applied both resolution enhancement fields and estimation wall-parallel velocity measurements shear stress pressure. analysis has been carried out with database turbulent open-channel flow friction Reynolds number $Re_{\tau}=180$ generated through direct numerical simulation. Coarse have...

Modal-decomposition techniques are computational frameworks based on data aimed at identifying a low-dimensional space for capturing dominant flow features: the so-called modes. We propose deep probabilistic-neural-network architecture learning minimal and near-orthogonal set of non-linear modes from high-fidelity turbulent-flow useful analysis, reduced-order modeling control. Our approach is β-variational autoencoders (β-VAEs) convolutional neural networks (CNNs), which enable extracting...

Deep reinforcement learning (DRL) has been applied to a variety of problems during the past decade, and provided effective control strategies in high-dimensional non-linear situations that are challenging traditional methods. Flourishing applications now spread out into field fluid dynamics, specifically active flow (AFC). In community AFC, encouraging results obtained two-dimensional chaotic conditions have raised interest study increasingly complex flows. this review, we first provide...

Turbulence is a complex phenomenon that has chaotic nature with multiple spatio-temporal scales, making predictions of turbulent flows challenging topic. Nowadays, an abundance high-fidelity databases can be generated by experimental measurements and numerical simulations, but obtaining such accurate data in full-scale applications currently not possible. This motivates utilising deep learning on subsets the available to reduce required cost reconstructing full flow applications. Here, we...

We introduce a reinforcement learning (RL) environment to design and benchmark control strategies aimed at reducing drag in turbulent fluid flows enclosed channel. The provides framework for computationally efficient, parallelized, high-fidelity simulations, ready interface with established RL agent programming interfaces. This allows both testing existing deep (DRL) algorithms against challenging task, advancing our knowledge of complex, physical system that has been major topic research...

Abstract The development of artificial intelligence (AI) as a field has impacted almost all aspects human life. More recently it found role in addressing developmental challenges, specifically the Sustainable Development Goals (SDGs). However, there are not enough systematic studies on analysis AI research towards SDGs. Therefore, this article attempts to bridge gap by identifying major bibliometric trends and concept‐evolution trajectories area applications for sustainable‐development...

This study proposes a newly developed deep-learning-based method to generate turbulent inflow conditions for spatially developing boundary layer (TBL) simulations. A combination of transformer and multiscale-enhanced super-resolution generative adversarial network is utilised predict velocity fields TBL at various planes normal the streamwise direction. Datasets direct numerical simulation (DNS) flat plate flow spanning momentum thickness-based Reynolds number, $Re_\theta =...

Physics is a field of science that has traditionally used the scientific method to answer questions about why natural phenomena occur and make testable models explain phenomena. Discovering equations, laws, principles are invariant, robust, causal been fundamental in physical sciences throughout centuries. Discoveries emerge from observing world and, when possible, performing interventions on system under study. With advent big data data-driven methods, fields equation discovery have...

Abstract Variational autoencoder architectures have the potential to develop reduced-order models for chaotic fluid flows. We propose a method learning compact and near-orthogonal using combination of β -variational transformer, tested on numerical data from two-dimensional viscous flow in both periodic regimes. The is trained learn latent representation velocity, transformer predict temporal dynamics latent-space. Using disentangled representations latent-space, we obtain more interpretable...

The world should redouble its efforts on the SDGs, not abandon them. Here's how to progress United Nations' agenda towards 2050.

Abstract High-resolution reconstruction of flow-field data from low-resolution and noisy measurements is interest due to the prevalence such problems in experimental fluid mechanics, where measurement are general sparse, incomplete noisy. Deep-learning approaches have been shown suitable for super-resolution tasks. However, a high number high-resolution examples needed, which may not be available many cases. Moreover, obtained predictions lack complying with physical principles, e.g. mass...

Turbulent boundary layers under adverse pressure gradients are studied using well-resolved large-eddy simulations (LES) with the goal of assessing influence streamwise pressure-gradient development. Near-equilibrium were characterized through Clauser parameter $\unicode[STIX]{x1D6FD}$ . In order to fulfil near-equilibrium conditions, free stream velocity was prescribed such that it followed a power-law distribution. The turbulence statistics pertaining cases constant value (extending up...

AbstractThree-dimensional effects in turbulent duct flows, i.e., sidewall boundary layers and secondary motions, are studied by means of direct numerical simulation (DNS). The spectral element code Nek5000 is used to compute flows with aspect ratios 1–7 (at Reb, c = 2800, Reτ, ≃ 180) ratio 1 5600, 330), streamwise-periodic boxes length 25h. total number grid points ranges from 28 145 million, the pressure gradient adjusted iteratively order keep same bulk Reynolds centreplane changing ratio....

In the present work, we analyze three commonly used methods to determine edge of pressure gradient turbulent boundary layers: two based on composite profiles, one by Chauhan et al. [“Criteria for assessing experiments in zero layers,” Fluid Dyn. Res. 41, 021404 (2009)] and Nickels [“Inner scaling wall-bounded flows subject large gradients,” J. Mech. 521, 217–239 (2004)], other condition vanishing mean velocity gradient. Additionally, a new method is introduced diagnostic plot concept...