Article
Chemistry, Physical
Michele Buzzicotti, Fabio Bonaccorso
Summary: The problem of classifying turbulent environments from partial observation is crucial for various fields, and can be approached using machine learning and Bayesian inference methods.
EUROPEAN PHYSICAL JOURNAL E
(2022)
Article
Multidisciplinary Sciences
E. Forero-Ortiz, G. Tirabassi, C. Masoller, A. J. Pons
Summary: This study successfully infers the connectivity and coupling strength of a network of coupled chaotic oscillators using the Kalman filter technique, even when oscillators are close to synchronization.
SCIENTIFIC REPORTS
(2021)
Article
Meteorology & Atmospheric Sciences
Mohamad Abed El Rahman Hammoud, Edriss S. Titi, Ibrahim Hoteit, Omar Knio
Summary: Generating high-resolution flow fields is crucial for engineering and climate science applications. We propose a physics-informed deep neural network (PI-DNN) to predict fine-scale flow fields using coarse-scale data. Numerical results show that the predictions of the PI-DNN are comparable to those obtained by dynamical downscaling.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2022)
Article
Energy & Fuels
Qiang Zheng, Xiaoguang Yin, Dongxiao Zhang
Summary: Li-ion battery is a complex physicochemical system with observable input and output variables, but also unobservable internal variables. Estimation for the unobservable states is crucial for battery management system. To achieve accuracy and efficiency in battery modeling, a data-driven surrogate integrating underlying physics is proposed. The surrogate, built upon operator networks, shows great potential for on-board scenarios, combining efficiency and accuracy, as well as offering an interface for model refinement.
JOURNAL OF ENERGY STORAGE
(2023)
Article
Meteorology & Atmospheric Sciences
Raju Attada, Ravi Kumar Kunchala, Hari Prasad Dasari, Sanikommu Sivareddy, Viswanadhapalli Yesubabu, Omar Knio, Ibrahim Hoteit
Summary: This study evaluates the performance of the WRF model in simulating the summer climate of the Arabian Peninsula, finding that the use of Spectral Nudging (SPN) method reduces biases and leads to more realistic simulations compared to the version without it (CTRL). SPN also captures the seasonal patterns of various climate features more accurately.
THEORETICAL AND APPLIED CLIMATOLOGY
(2021)
Article
Chemistry, Physical
Amanda A. Howard, Tong Yu, Wei Wang, Alexandre M. Tartakovsky
Summary: Vanadium redox flow batteries have the potential to store large amounts of energy cheaply and efficiently. We have developed a multifidelity model to predict the charge-discharge curve of a VRFB, which shows good agreement with experimental results and significant improvements over existing models.
JOURNAL OF POWER SOURCES
(2022)
Article
Thermodynamics
S. Hanrahan, M. Kozul, R. D. Sandberg
Summary: This paper introduces the application of physics-informed neural networks (PINNs) and their advantages in simulating fluid dynamics problems. By combining deep neural networks with partial differential equations, PINNs can produce reliable predictions with sparse training data. The authors demonstrate the accuracy and robustness of PINNs in simulating different flow problems through simulations of adverse-pressure-gradient boundary layer and periodic hill problems.
INTERNATIONAL JOURNAL OF HEAT AND FLUID FLOW
(2023)
Article
Computer Science, Artificial Intelligence
John J. Molina, Kenta Ogawa, Takashi Taniguchi
Summary: This article introduces a probabilistic Stokes flow framework developed using physics informed Gaussian processes, which can be used to solve both forward/inverse flow problems with missing and/or noisy data. It automatically selects physically meaningful velocity fields without the need to explicitly solve the Poisson equation for the pressure field. Applying this method in the analysis of experimental data can handle common noisy/missing data.
MACHINE LEARNING-SCIENCE AND TECHNOLOGY
(2023)
Article
Physics, Multidisciplinary
Jurriaan W. R. Peeters
Summary: A direct link between Stanton number and scalar spectrum is established using the phenomenological theory of turbulence, showing the effect of different scales of motion on heat transfer. The study reproduces two important observations in literature by considering only the viscous inertial and diffusive range of the scalar spectrum.
PHYSICAL REVIEW LETTERS
(2023)
Article
Computer Science, Information Systems
Muhammad Rafiq, Ghazala Rafiq, Gyu Sang Choi
Summary: Solving parametric partial differential equations using artificial intelligence allows for faster convergence and higher accuracy compared to traditional numerical solvers.
Article
Mechanics
R. Laubscher
Summary: In this study, single- and segregated-network physics-informed neural network (PINN) architectures were applied to predict momentum, species, and temperature distributions in a dry air humidification problem. It was found that the segregated-network PINN approach resulted in significantly lower losses compared to the single-network PINN architecture, showcasing the importance of segregated approach. The PINN models produced accurate results for temperature and velocity profiles, but there is still room for improvement in the species mass fraction predictions.
Article
Mathematics, Applied
Salvatore Cuomo, Mariapia De Rosa, Fabio Giampaolo, Stefano Izzo, Vincenzo Schiano Di Cola
Summary: In recent years, Scientific Machine Learning (SciML) methods, particularly Physics-Informed Neural Networks (PINNs), have become popular for solving non-linear partial differential equations (PDEs). This paper numerically tackles the groundwater flow equations using a PINN approach, approximating the Dirac distribution and analyzing its computational ability in higher-dimensional cases. The effectiveness of PINNs is demonstrated through numerical experiments in hydrological applications, comparing the results with the Finite Difference Method (FDM) and highlighting the advantages of PINNs in solving PDEs without discretization.
COMPUTERS & MATHEMATICS WITH APPLICATIONS
(2023)
Article
Engineering, Multidisciplinary
Shuaijun Lv, Daolun Li, Wenshu Zha, Luhang Shen, Yan Xing
Summary: This paper proposes a physics-informed residual network (PIResNet) to solve the single-phase seepage equation without labeled data. It adds physical constraints to the neural network, constructs the loss function based on the residuals of the discretized seepage equation, and embeds the boundary conditions as hard constraints. PIResNet has a simple network structure, fast convergence, and easy optimization.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2024)
Article
Environmental Sciences
QiZhi He, Alexandre M. Tartakovsky
Summary: A discretization-free approach based on the PINN method is proposed to solve the coupled advection-dispersion equation and Darcy flow equation, using deep neural networks for approximating hydraulic conductivity, hydraulic head, and concentration fields, leading to improved accuracy in computations. The PINN method demonstrates high accuracy for both forward and backward ADEs, with the inclusion of concentration measurements significantly enhancing solution accuracy.
WATER RESOURCES RESEARCH
(2021)
Article
Engineering, Multidisciplinary
John M. Hanna, Jose V. Aguado, Sebastien Comas-Cardona, Ramzi Askri, Domenico Borzacchiello
Summary: This paper proposes a machine learning framework for simulating two-phase flow in porous media. The algorithm, based on Physics-informed neural networks (PINN), introduces a novel residual-based adaptive approach. The results demonstrate that this method can accurately capture moving flow fronts compared to traditional methods.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2022)
Article
Mechanics
Xiangjun Wang, Minping Wang, Luca Biferale
Summary: This study investigates the accelerations of tracer and light particles in compressible homogeneous isotropic turbulence. The results show that the characteristics of acceleration vary for tracer particles and bubbles, and are closely related to the flow structures.
JOURNAL OF FLUID MECHANICS
(2022)
Article
Multidisciplinary Sciences
Andrea Mazzino, Marco Edoardo Rosti
Summary: This study provides a numerical validation of a phenomenological theory for characterizing the statistical properties of a turbulent puff. The theory covers both the bulk properties and the scaling laws for velocity and temperature differences in different scales. The study also discovers a new mechanism dominated by buoyancy in turbulent fluctuations and provides a theoretical explanation for it.
PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY A-MATHEMATICAL PHYSICAL AND ENGINEERING SCIENCES
(2022)
Article
Multidisciplinary Sciences
A. Alexakis, L. Biferale
Summary: In this study, the authors numerically investigate the effect of a parameter lambda on the energy cascade direction in the Navier-Stokes equations. They find that as lambda approaches a critical value lambda(c), the kinetic energy diverges and the energy spectrum exhibits a larger bottleneck. Furthermore, they observe an increase in the amplitudes of both the forward heterochiral flux and the inverse homochiral flux, while their difference remains fixed. They also observe a reduction in intermittency as lambda approaches lambda(c).
PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY A-MATHEMATICAL PHYSICAL AND ENGINEERING SCIENCES
(2022)
Article
Mechanics
Lokahith Agasthya, Patricio Clark Di Leoni, Luca Biferale
Summary: The study applies nudging technique to reconstruct temperature fields in a Rayleigh-Benard convection system at varying turbulence levels, evaluating the success in reconstructing flow fields and transitioning to synchronization. Additionally, the study examines the statistical reproduction of system dynamics and accurate prediction of heat transfer properties.
Article
Mechanics
M. Cavaiola, S. Olivieri, J. Guerrero, A. Mazzino, M. E. Rosti
Summary: State-of-the-art direct numerical simulations were used to study the role of barriers in the airborne spread of virus-containing droplets. The study found that barriers have nontrivial dynamical effects on the final reach of virus-containing droplets and may not always be beneficial. These conclusions depend on the relative humidity of the ambient condition.
Correction
Physics, Multidisciplinary
Simona Colabrese, Kristian Gustavsson, Antonio Celani, Luca Biferale
PHYSICAL REVIEW LETTERS
(2022)
Article
Physics, Fluids & Plasmas
Francesco Borra, Luca Biferale, Massimo Cencini, Antonio Celani
Summary: This study focuses on a model of two competing microswimming agents engaged in a pursue-evasion task within a low-Reynolds-number environment. The agents, with limited maneuverability and partial information about the opponent's position and motion, are trained using adversarial reinforcement learning to overcome partial observability and discover increasingly complex sequences of moves.
PHYSICAL REVIEW FLUIDS
(2022)
Article
Mechanics
Stefano Olivieri, Andrea Mazzino, Marco E. Rosti
Summary: This study investigates the mechanical behavior of finite-size, elastic and inertial fibers in a homogeneous and isotropic turbulent flow. The results reveal a robust turbulence modulation mechanism primarily controlled by the mass fraction of the suspension, with minor influence from the fiber's bending stiffness. The study also explores the different flapping states and scaling laws of the fibers' maximum curvature. Furthermore, clustering and preferential alignment of fibers within the flow are examined, highlighting the role of inertia and elasticity.
JOURNAL OF FLUID MECHANICS
(2022)
Article
Thermodynamics
Gabriele Casciaro, Francesco Ferrari, Mattia Cavaiola, Andrea Mazzino
Summary: The issue of the accuracy of wind speed/power forecasts is becoming increasingly important. A novel Ensemble Model Output Statistics (EMOS) strategy is proposed to improve wind speed/power forecasts by considering nonlinear relationships and conditioning variables. It shows a net improvement compared to ordinary EMOS strategies.
ENERGY CONVERSION AND MANAGEMENT
(2022)
Article
Computer Science, Interdisciplinary Applications
Patricio Clark Di Leoni, Lu Lu, Charles Meneveau, George Em Karniadakis, Tamer A. Zaki
Summary: This study investigates the prediction of linear evolution of instability waves in high-speed boundary layers using neural operators. The design of DeepOnet is extended to ensure accurate and robust predictions, and data assimilation is also performed. DeepOnet is trained to take upstream disturbance and downstream location as inputs and provide perturbation field downstream as output, approximating the linearized and parabolized Navier-Stokes operator. The trained DeepOnet can perform fast and accurate predictions of downstream disturbances within the training frequency range.
JOURNAL OF COMPUTATIONAL PHYSICS
(2023)
Article
Mechanics
Lokahith Agasthya, Andreas Bartel, Luca Biferale, Matthias Ehrhardt, Federico Toschi
Summary: This article investigates the complex interactions between non-isothermal particles suspended in a fluid and provides a numerical study that demonstrates the control of thermal convection through pure Lagrangian forcing.
JOURNAL OF FLUID MECHANICS
(2022)
Article
Chemistry, Physical
Patricio Clark Di Leoni, Lokahith Agasthya, Michele Buzzicotti, Luca Biferale
Summary: We study the use of Physics-Informed Neural Networks (PINNs) for reconstructing turbulent Rayleigh-Benard flows using temperature information only. We perform a quantitative analysis of the reconstruction quality at different levels of low-pass filtered information and turbulent intensities. Our results show that PINNs achieve high precision reconstruction comparable to nudging at low Rayleigh numbers. At high Rayleigh numbers, PINNs outperform nudging and achieve satisfactory reconstruction of velocity fields with high spatial and temporal density of temperature data. However, the performance of PINNs deteriorates when data becomes sparse, both in terms of point-to-point errors and in a statistical sense, as observed in probability density functions and energy spectra.
EUROPEAN PHYSICAL JOURNAL E
(2023)
Article
Chemistry, Physical
Sofia Angriman, Pablo Cobelli, Pablo D. Mininni, Martin Obligado, Patricio Clark Di Leoni
Summary: When modeling turbulent flows, it is often difficult to obtain or implement information on forcing terms or boundary conditions. This study introduces a method based on physics-informed neural networks to assimilate given conditions into turbulent states. Examples of different statistical conditions are provided for state preparation, inspired by experimental and atmospheric problems. Two ways of scaling the resolution of prepared states are demonstrated: using multiple and parallel neural networks, or leveraging the power of specialized numerical solvers through synchronization-based data assimilation technique called nudging.
EUROPEAN PHYSICAL JOURNAL E
(2023)
Article
Mechanics
Damiano Capocci, Perry L. Johnson, Sean Oughton, Luca Biferale, Moritz Linkmann
Summary: In this study, the relative contributions of different physical mechanisms to the energy cascade and helicity transfer in turbulence are quantified. It is found that scale-local vortex flattening and twisting dominate the helicity transfer, accounting for approximately 50% of the mean flux. A new exact relation between these effects is derived, showing the dominance of vortex flattening over twisting. The remaining 50% of the mean flux is attributed to multi-scale vortex flattening, twisting, and entangling.
JOURNAL OF FLUID MECHANICS
(2023)
Article
Physics, Fluids & Plasmas
R. A. Heinonen, L. Biferale, A. Celani, M. Vergassola
Summary: In many practical scenarios, flying insects face the challenge of searching for emitted cues in turbulent environments. In this study, a partially observable Markov decision process is used to model this search problem, and the Perseus algorithm is employed to compute near-optimal strategies for minimizing arrival time. The computed strategies outperform several heuristic strategies and provide insights into the search difficulty and robustness in different environmental conditions.
