Related references
Note: Only part of the references are listed.
Article
Computer Science, Interdisciplinary Applications
Amirhossein Arzani et al.
Summary: Physics-informed neural networks (PINNs) have become popular for scientific machine learning and differential equation modeling. This study introduces boundary-layer PINN (BL-PINN), which treats thin boundary layers as singular perturbation problems. By incorporating classical perturbation theory, BL-PINN employs different parallel PINN networks to approximate the boundary layer problem in both inner and outer regions. BL-PINN outperforms traditional PINN and other extensions such as XPINN in various benchmark problems, providing accurate solutions.
JOURNAL OF COMPUTATIONAL PHYSICS
(2023)
Editorial Material
Energy & Fuels
Piyang Liu et al.
Summary: Reservoir history matching refers to continuously adjusting the parameters of the reservoir model to match the historical observation data, which is essential for making forecasts based on the reservoir model. Automatic history matching has made significant progress for practical applications due to advances in optimization theory and machine learning algorithms. This article summarizes the existing automatic history matching methods, categorizing them as model-driven or surrogate-driven based on whether the reservoir simulator needs to be run. The basic principles and limitations of these methods in practical applications are outlined, followed by a discussion on the future trends of reservoir automatic history matching.
ADVANCES IN GEO-ENERGY RESEARCH
(2023)
Article
Computer Science, Artificial Intelligence
Wenshu Zha et al.
Summary: This letter presents 3D-PDE-Net, a neural network designed for solving three-dimensional partial differential equations (PDEs). It introduces a mathematical derivation of a three-dimensional convolution kernel that can approximate any order differential operator, and utilizes this theory to construct 3D-PDE-Net. Experimental results demonstrate that 3D-PDE-Net can achieve accurate solutions with few training samples, and it is highly significant in solving linear and nonlinear unsteady PDEs.
NEURAL COMPUTATION
(2022)
Article
Mechanics
Qiuwan Du et al.
Summary: In this paper, a convolutional neural network framework is established based on the deep learning method for airfoil design optimization and performance prediction. The experimental results show that the framework has significant advantages such as good robustness, great convergence, fast computation speed, and high prediction accuracy.
Article
Mechanics
Yu Liu et al.
Summary: This study investigated the spray characteristics of biodiesel and polyoxymethylene dimethyl ethers (PODE) under different injection pressures. It was found that biodiesel blends have a greater spray tip penetration compared to diesel blends, and the addition of PODE brings the spray characteristics of biodiesel closer to pure diesel. An excellent predictive model was also developed using computer technology.
Article
Energy & Fuels
Nanzhe Wang et al.
Summary: This study extends a theory-guided convolutional neural network (TgCNN) framework as a surrogate model for subsurface flows, incorporating physical constraints for better accuracy and generalizability. The TgCNN surrogate, combined with a genetic algorithm, is used for well placement optimization, achieving satisfactory performance even with limited data and varying well numbers. The results demonstrate that the TgCNN surrogate-based optimization strategy is accurate and efficient, outperforming direct simulator-based optimization methods and maintaining stability under geologic uncertainty.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2022)
Article
Energy & Fuels
Luhang Shen et al.
Summary: This paper proposes a physics-informed method to solve a porous flow equation using neural networks, achieving higher accuracy and allowing for the inversion of permeability from measurable data.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2022)
Article
Mechanics
Sandeep Pandey et al.
Summary: In this study, a combined convolutional autoencoder-recurrent neural network model is proposed to analyze and forecast the dynamics and low-order statistics of local convective heat flux in a two-dimensional turbulent Rayleigh-Benard convection flow. The model achieves good accuracy in statistical properties and is able to reproduce the intermittent plume-mixing dynamics and probability density function of the convective heat flux.
Article
Physics, Fluids & Plasmas
Marios Mattheakis et al.
Summary: The research introduces a Hamiltonian neural network to solve differential equations governing dynamical systems without using ground truth data, optimizing solely based on predicted functions to improve accuracy. Error analysis indicates numerical errors depend on network performance.
Article
Engineering, Multidisciplinary
Sina Amini Niaki et al.
Summary: The study introduces a Physics-Informed Neural Network (PINN) to simulate the thermal-chemical evolution of a composite material curing in an autoclave. By optimizing deep neural network (DNN) parameters using a physics-based loss function, the research solves coupled differential equations, designs a PINN with two disconnected subnetworks, and develops a sequential training algorithm. The approach incorporates explicit discontinuities at the composite-tool interface and enforces known physical behavior in the loss function to enhance solution accuracy.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2021)
Article
Mechanics
Sandeep Reddy Bukka et al.
Summary: This paper introduces two deep learning-based hybrid data-driven reduced-order models for predicting unsteady fluid flows. The first model (POD-RNN) projects time series data to a low-dimensional space, while the second model (CRAN) extracts low-dimensional features using convolutional neural networks. Through benchmark problems, it is found that the CRAN model outperforms the POD-RNN model in complex flow scenarios.
Article
Mechanics
Suraj Pawar et al.
Summary: The study introduces a physics-guided machine learning framework to improve the accuracy of data-driven predictive engines by adding features in intermediate layers to emphasize physical importance. By addressing generalizability concerns, the results suggest that the proposed feature enhancement approach can be effectively used in many scientific machine learning applications.
Article
Mechanics
L. M. Yang et al.
Summary: A high-order gas kinetic flux solver (GKFS) is developed within the framework of the high-order finite volume (FV) method for simulation of two-dimensional incompressible flows, which evaluates inviscid and viscous fluxes simultaneously to achieve higher efficiency and accuracy compared to conventional methods.
Article
Energy & Fuels
Li Daolun et al.
Summary: This paper introduces an improved physics-constrained PDE solution method, which incorporates gradient models to enhance the accuracy of neural networks in approximating PDEs. This approach solves complex partial differential equation problems, such as those with source and sink terms, by adding special neurons in the hidden layer.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2021)
Article
Engineering, Multidisciplinary
Minglang Yin et al.
Summary: The study utilizes physics-informed neural networks (PINNs) to infer properties of biological materials using synthetic data, successfully applied to extract permeability and viscoelastic modulus from thrombus deformation data. The results demonstrate that PINNs can infer material properties from noisy synthetic data and have great potential for inferring these properties from experimental multi-modality and multi-fidelity data.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2021)
Article
Computer Science, Interdisciplinary Applications
Xiaowei Jin et al.
Summary: In the past 50 years, significant progress has been made in solving Navier-Stokes equations using various numerical methods, but challenges still exist in incorporating multi-fidelity data seamlessly and mesh generation for complex industrial applications. Physics-informed neural networks (PINNs) are employed to directly encode governing equations into deep neural networks, overcoming some limitations for simulating incompressible laminar and turbulent flows. The Navier-Stokes flow nets (NSFnets) are developed using velocity-pressure (VP) and vorticity-velocity (VV) formulations, showing promise in accuracy and convergence rate for benchmark problems as well as turbulent channel flows.
JOURNAL OF COMPUTATIONAL PHYSICS
(2021)
Article
Mechanics
Zhiwen Deng et al.
Summary: This paper focuses on the optimal sensor placement strategy based on a deep neural network for turbulent flow recovery within the data assimilation framework of the ensemble Kalman filter. The results demonstrate the effectiveness and robustness of the proposed strategy, showing that RANS models with EnKF augmentation were substantially improved over their original counterparts. The study concludes that the DNN-based OSP with the selection of the five most sensitive sensors can efficiently reduce the number of sensors while achieving similar or better assimilated performance.
Article
Mechanics
Chen Kong et al.
Summary: Through a data-driven approach, a deep learning model based on CNN successfully learned the relationship between the velocity field and the wall pressure in the isolator, and accurately reconstructed the velocity field under different Mach numbers and backpressures.
Article
Mechanics
Romit Maulik et al.
Summary: This study demonstrates that using a combination of convolutional autoencoders (CAEs) and recurrent neural networks effectively overcomes the limitations of the POD-Galerkin technique in advection-dominated nonlinear PDEs. Truncated systems with only a few latent space dimensions can accurately reproduce complex fluid dynamics phenomena, and parameter information can be easily embedded to detect trends in system evolution.
Article
Multidisciplinary Sciences
Sifan Wang et al.
Summary: The physics-informed DeepONets framework can effectively learn the solution operator of arbitrary PDEs even in the absence of paired training data. It is able to predict the solution of various parametric PDEs up to three orders of magnitude faster compared to conventional solvers, setting a new paradigm for modeling and simulation of nonlinear and nonequilibrium processes in science and engineering.
Article
Mechanics
S. B. Leask et al.
Summary: Recent deep learning extensions in Koopman theory have enabled the use of Deep Koopman networks to automatically identify independent modes in nonlinear dynamical systems observed optically, demonstrating flexibility and interpretability in system data collection.
Article
Mechanics
Hao Xu et al.
Summary: Data-driven methods have recently made great progress in discovering partial differential equations from spatial-temporal data. A new framework combining neural network, genetic algorithm, and stepwise methods is proposed to address challenges including sparse noisy data and incomplete library. The proposed algorithm is able to discover parametric PDEs with spatially or temporally varying coefficients on various equations.
Article
Mathematics, Applied
Rong Yang et al.
Summary: This paper focuses on the long-term asymptotic stability of global weak solutions for 3D incompressible Navier-Stokes equations with nonlinear damping, specifically looking at the convergence of solutions between Navier-Stokes systems and their corresponding generalized elliptic equations.
APPLIED MATHEMATICS LETTERS
(2021)
Article
Engineering, Multidisciplinary
Ehsan Haghighat et al.
Summary: This study presents the application of Physics Informed Neural Networks (PINN) in solid mechanics, improving accuracy and convergence with a multi-network model and Isogeometric Analysis. The study demonstrates the importance of honoring physics in improving robustness and highlights the potential application of PINN in sensitivity analysis and surrogate modeling.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2021)
Article
Engineering, Multidisciplinary
Luning Sun et al.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2020)
Article
Computer Science, Interdisciplinary Applications
Yating Wang et al.
JOURNAL OF COMPUTATIONAL PHYSICS
(2020)
Article
Multidisciplinary Sciences
Maziar Raissi et al.
Article
Engineering, Multidisciplinary
Ameya D. Jagtap et al.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2020)
Article
Engineering, Multidisciplinary
Xuhui Meng et al.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2020)
Article
Mechanics
Suraj Pawar et al.
Article
Mechanics
Chenyue Xie et al.
Article
Energy & Fuels
Li Daolun et al.
PETROLEUM EXPLORATION AND DEVELOPMENT
(2020)
Article
Computer Science, Interdisciplinary Applications
M. Raissi et al.
JOURNAL OF COMPUTATIONAL PHYSICS
(2019)
Article
Computer Science, Interdisciplinary Applications
Yinhao Zhu et al.
JOURNAL OF COMPUTATIONAL PHYSICS
(2019)
Article
Environmental Sciences
Shaoxing Mo et al.
WATER RESOURCES RESEARCH
(2019)
Article
Computer Science, Interdisciplinary Applications
Justin Sirignano et al.
JOURNAL OF COMPUTATIONAL PHYSICS
(2018)
Article
Mechanics
R. Maulik et al.
Article
Physics, Mathematical
Chunyan Wang et al.
REPORTS ON MATHEMATICAL PHYSICS
(2017)
Article
Engineering, Petroleum
Daolun Li et al.