Physics-Guided, Physics-Informed, and Physics-Encoded Neural Networks and Operators in Scientific Computing: Fluid and Solid Mechanics

Journal of Computing and Information Science in Engineering, Journal Year: 2024, Volume and Issue: 24(4)

Published: Jan. 8, 2024

Abstract Advancements in computing power have recently made it possible to utilize machine learning and deep push scientific forward a range of disciplines, such as fluid mechanics, solid materials science, etc. The incorporation neural networks is particularly crucial this hybridization process. Due their intrinsic architecture, conventional cannot be successfully trained scoped when data are sparse, which the case many engineering domains. Nonetheless, provide foundation respect physics-driven or knowledge-based constraints during training. Generally speaking, there three distinct network frameworks enforce underlying physics: (i) physics-guided (PgNNs), (ii) physics-informed (PiNNs), (iii) physics-encoded (PeNNs). These methods advantages for accelerating numerical modeling complex multiscale multiphysics phenomena. In addition, recent developments operators (NOs) add another dimension these new simulation paradigms, especially real-time prediction systems required. All models also come with own unique drawbacks limitations that call further fundamental research. This study aims present review four (i.e., PgNNs, PiNNs, PeNNs, NOs) used state-of-the-art architectures applications reviewed, discussed, future research opportunities presented terms improving algorithms, considering causalities, expanding applications, coupling solvers.

Language: Английский

---

A comprehensive review of advances in physics-informed neural networks and their applications in complex fluid dynamics

Physics of Fluids, Journal Year: 2024, Volume and Issue: 36(10)

Published: Oct. 1, 2024

Physics-informed neural networks (PINNs) represent an emerging computational paradigm that incorporates observed data patterns and the fundamental physical laws of a given problem domain. This approach provides significant advantages in addressing diverse difficulties field complex fluid dynamics. We thoroughly investigated design model architecture, optimization convergence rate, development modules for PINNs. However, efficiently accurately utilizing PINNs to resolve dynamics problems remain enormous barrier. For instance, rapidly deriving surrogate models turbulence from known characterizing flow details multiphase fields present substantial difficulties. Additionally, prediction parameters multi-physics coupled models, achieving balance across all scales multiscale modeling, developing standardized test sets encompassing dynamic are urgent technical breakthroughs needed. paper discusses latest advancements their potential applications dynamics, including turbulence, flows, multi-field flows. Furthermore, we analyze challenges face these outline future trends growth. Our objective is enhance integration deep learning facilitating resolution more realistic problems.

Language: Английский

---

Risks and uncertainties in carbon capture, transport, and storage projects: A comprehensive review

Gas Science and Engineering, Journal Year: 2023, Volume and Issue: 119, P. 205117 - 205117

Published: Sept. 21, 2023

Language: Английский

---

A comparative analysis of super-resolution techniques for enhancing micro-CT images of carbonate rocks

Applied Computing and Geosciences, Journal Year: 2023, Volume and Issue: 20, P. 100143 - 100143

Published: Nov. 14, 2023

High-resolution digital rock micro-CT images captured from a wide field of view are essential for various geosystem engineering and geoscience applications. However, the resolution these is often constrained by capabilities scanners. To overcome this limitation achieve superior image quality, advanced deep learning techniques have been used. This study compares four different super-resolution techniques, including convolutional neural network (SRCNN), efficient sub-pixel networks (ESPCN), enhanced residual (EDRN), generative adversarial (SRGAN) to enhance obtained heterogeneous porous media. Our investigation employs dataset consisting 5,000 acquired highly carbonate rock. The performance each algorithm evaluated based on its accuracy reconstruct pore geometry connectivity, grain-pore edge sharpness, preservation petrophysical properties, such as porosity. findings indicate that EDRN outperforms other in terms peak signal-to-noise ratio (PSNR) structural similarity (SSIM) index, increased nearly 4 dB 17%, respectively, compared bicubic interpolation. Furthermore, SRGAN exhibits technqiues learned perceptual patch (LPIPS) index porosity error. shows 30% reduction LPIPS results provide deeper insights into practical applications domain media characterizations, facilitating selection optimal CNN-based methodologies.

Language: Английский

---

Uncertainty Quantification in CO2 Trapping Mechanisms: A Case Study of PUNQ-S3 Reservoir Model Using Representative Geological Realizations and Unsupervised Machine Learning

Energies, Journal Year: 2024, Volume and Issue: 17(5), P. 1180 - 1180

Published: March 1, 2024

Evaluating uncertainty in CO2 injection projections often requires numerous high-resolution geological realizations (GRs) which, although effective, are computationally demanding. This study proposes the use of representative (RGRs) as an efficient approach to capture range full set while reducing computational costs. A predetermined number RGRs is selected using integrated unsupervised machine learning (UML) framework, which includes Euclidean distance measurement, multidimensional scaling (MDS), and a deterministic K-means (DK-means) clustering algorithm. In context intricate 3D aquifer storage model, PUNQ-S3, these algorithms utilized. The UML methodology selects five from pool 25 possibilities (20% total), taking into account reservoir quality index (RQI) static parameter reservoir. To determine credibility RGRs, their simulation results scrutinized through application Kolmogorov–Smirnov (KS) test, analyzes distribution output. this assessment, 40 wells cover entire alongside set. end-point indicate that structural, residual, solubility trapping within follow same distribution. Simulating GRs over 200 years, involving 10 years injection, reveals consistently similar patterns, with average value Dmax 0.21 remaining lower than Dcritical (0.66). Using methodology, expenses related scenario testing development planning for reservoirs presence uncertainties can be substantially reduced.

Language: Английский

---

Performance of physical-informed neural network (PINN) for the key parameter inference in Langmuir turbulence parameterization scheme

Acta Oceanologica Sinica, Journal Year: 2024, Volume and Issue: 43(5), P. 121 - 132

Published: May 1, 2024

Language: Английский

---

A physics-informed neural network framework for multi-physics coupling microfluidic problems

Computers & Fluids, Journal Year: 2024, Volume and Issue: 284, P. 106421 - 106421

Published: Sept. 4, 2024

Language: Английский

---

Physics informed neural network for forward and inverse multispecies contaminant transport with variable parameters

Journal of Hydrology, Journal Year: 2025, Volume and Issue: unknown, P. 132977 - 132977

Published: Feb. 1, 2025

Language: Английский

---

Learning High-Dimensional Chaos Based on an Echo State Network with Homotopy Transformation

Mathematics, Journal Year: 2025, Volume and Issue: 13(6), P. 894 - 894

Published: March 7, 2025

Learning high-dimensional chaos is a complex and challenging problem because of its initial value-sensitive dependence. Based on an echo state network (ESN), we introduce homotopy transformation in topological theory to learn chaos. On the premise maintaining basic properties, our model can obtain key features for learning through continuous between different activation functions, achieving optimal balance nonlinearity linearity enhance generalization capability model. In experimental part, choose Lorenz system, Mackey–Glass (MG) Kuramoto–Sivashinsky (KS) system as examples, verify superiority by comparing it with other models. For some systems, prediction error be reduced two orders magnitude. The results show that addition improve modeling ability spatiotemporal chaotic this demonstrates potential application dynamic time series analysis.

Language: Английский

---

Physics-informed neural networks for Kelvin–Helmholtz instability with spatiotemporal and magnitude multiscale

Physics of Fluids, Journal Year: 2025, Volume and Issue: 37(3)

Published: March 1, 2025

Prediction of Kelvin–Helmholtz instability (KHI) is crucial across various fields, requiring extensive high-fidelity data. However, experimental data are often sparse and noisy, while simulated may lack credibility due to discrepancies with real-world configurations parameters. This underscores the need for field reconstruction parameter inference from sparse, noisy data, which constitutes inverse problems. Based on physics-informed neural networks (PINNs), KH-PINN framework established in this work solve problems KHI flows. By incorporating governing physical equations, reconstructs continuous flow fields infer unknown transport parameters observed The two-dimensional unsteady incompressible flows both constant variable densities studied. To our knowledge, one first few applications PINNs densities. address spatiotemporal multiscale issue enhance accuracy small-scale structures, embedding (ME) strategy adopted. magnitude small-magnitude velocities, critical problems, small-velocity amplification (SVA) proposed. results demonstrate that can accurately reconstruct complex, evolving vortices a broad range Reynolds numbers. Additionally, energy-decaying entropy-increasing curves obtained. effectiveness ME SVA validated through comparative studies, anti-noise few-shot learning capabilities also validated. code available at https://github.com/CAME-THU/KH-PINN.

Language: Английский

---