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
Physics, Mathematical
Colby L. Wight, Jia Zhao
Summary: This paper focuses on using deep neural networks to design an improved Physics Informed Neural Network (PINN) for automatically solving the Allen-Cahn and Cahn-Hilliard equations. Various techniques and sampling strategies are proposed to enhance the efficiency and accuracy of the PINN in solving phase field equations, allowing for a wider applicability to a broader class of PDE problems without restriction on the explicit form of the PDEs.
COMMUNICATIONS IN COMPUTATIONAL PHYSICS
(2021)
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
Engineering, Multidisciplinary
Kevin Linka, Amelie Schafer, Xuhui Meng, Zongren Zou, George Em Karniadakis, Ellen Kuhl
Summary: Understanding real-world dynamical phenomena is challenging, and machine learning has become the go-to technology for analyzing and making decisions based on these phenomena. However, traditional neural networks often ignore the fundamental laws of physics and fail to make accurate predictions. In this study, the combination of neural networks, physics informed modeling, and Bayesian inference is used to integrate data, physics, and uncertainties, improving the predictive potential of neural network models.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2022)
Article
Multidisciplinary Sciences
Umit H. Coskun, Bilgehan Sel, Brad Plaster
Summary: This paper proposes a new method for accessing magnetic field components in experimentally inaccessible regions. It uses non-disruptive magnetic field measurements on a surface enclosing the experimental region and solves a set of partial differential equations with physics-informed neural networks. The method is benchmarked against an older method and is particularly useful for experiments requiring precise determination of magnetic field components.
SCIENTIFIC REPORTS
(2022)
Article
Mechanics
Mario De Florio, Enrico Schiassi, Barry D. Ganapol, Roberto Furfaro
Summary: This study accurately solves a thermal creep flow problem in a plane channel using Physics-Informed Neural Networks. By developing a specific framework that utilizes Constrained Expressions based on the Theory of Functional Connections, the study demonstrates that accurate solutions can be achieved with fast training times using shallow neural networks, such as Chebyshev NN and Legendre NN. The results show that the approach is effective in matching benchmarks and providing accurate solutions.
Article
Engineering, Multidisciplinary
Nilgun Guler Bayazit
Summary: This paper introduces a two-stage physics informed neural network, which significantly improves the prediction accuracy of PDE solutions by training each stage separately and not backpropagating the gradients to the first stage.
ENGINEERING SCIENCE AND TECHNOLOGY-AN INTERNATIONAL JOURNAL-JESTECH
(2023)
Article
Engineering, Multidisciplinary
Minglang Yin, Xiaoning Zheng, Jay D. Humphrey, George Em Karniadakis
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
Mathematics, Applied
Zhiwei Gao, Liang Yan, Tao Zhou
Summary: In this work, an adaptive strategy called FI-PINNs is proposed for solving PDE problems. The strategy defines failure probability based on the residual and improves numerical accuracy by adding more training points. Similar to adaptive finite element methods, the approach uses failure probability as a posterior error indicator to generate new training points.
SIAM JOURNAL ON SCIENTIFIC COMPUTING
(2023)
Article
Mechanics
Hamidreza Eivazi, Mojtaba Tahani, Philipp Schlatter, Ricardo Vinuesa
Summary: This article introduces the application of Physics-informed neural networks (PINNs) in solving and identifying partial differential equations. By applying PINNs to solve boundary layer problems of the Navier-Stokes equations and simulate various turbulent flow cases, it is demonstrated that PINNs have good applicability for both laminar and turbulent flows.
Article
Computer Science, Artificial Intelligence
Xu Liu, Xiaoya Zhang, Wei Peng, Weien Zhou, Wen Yao
Summary: This paper introduces a physics-informed neural network (PINN) with a new Reptile initialization (NRPINN) for solving scientific computing problems. The NRPINN efficiently solves partial differential equations (PDEs) by sampling tasks from parameterized PDEs and adjusting the penalty term of the loss function.
NEURAL COMPUTING & APPLICATIONS
(2022)
Article
Energy & Fuels
Jincheng Zhang, Xiaowei Zhao
Summary: This study develops a method that predicts spatiotemporal wind fields by combining LIDAR measurements and flow physics. A deep neural network incorporating the Navier-Stokes equations is trained using sparse LIDAR measurements, enabling the prediction of wind fields in the whole domain. The proposed method shows promising results in accurately predicting flow velocity under various scenarios using only sparse line-of-sight wind speed measurements.
Article
Energy & Fuels
Jincheng Zhang, Xiaowei Zhao
Summary: By combining LIDAR measurements and flow physics, this study developed a method to predict the spatiotemporal wind field, achieving promising results in various scenarios with only sparse LIDAR measurements. The approach utilized a deep neural network incorporating the Navier-Stokes equations, allowing for accurate predictions of wind patterns not present in the training dataset.
Article
Environmental Sciences
Libin Du, Zhengkai Wang, Zhichao Lv, Lei Wang, Dongyue Han
Summary: This paper proposes an underwater acoustic field prediction method based on a physics-informed neural network. By introducing a loss function incorporating physical constraints, the method can effectively predict the distribution of two-dimensional and three-dimensional underwater sound fields while ensuring clear physical significance of the trained model.
FRONTIERS IN MARINE SCIENCE
(2023)
Article
Computer Science, Interdisciplinary Applications
Huayan Pu, Bo Tan, Jin Yi, Shujin Yuan, Jinglei Zhao, Ruqing Bai, Jun Luo
Summary: This paper proposes a novel method based on physical-informed neural networks (PINN) to calculate the performance of the permanent magnet coupler (PMC). Experimental results demonstrate that PINN outperforms traditional calculation methods regarding feasibility, validity, and accuracy, and provides a new approach for optimizing PMC structure design.
ENGINEERING WITH COMPUTERS
(2023)
Article
Mechanics
Yao Xiao, Zhong Zeng, Liangqi Zhang, Jingzhu Wang, Yiwei Wang, Hao Liu, Chenguang Huang
Summary: In this paper, a spectral element-based phase field method is proposed for solving the Navier-Stokes/Cahn-Hilliard equations for incompressible two-phase flows. The method effectively tackles the challenge posed by the high-order nonlinear term of the Cahn-Hilliard equation by employing the Newton-Raphson method. The decoupling of the Navier-Stokes equations using a time-stepping scheme improves the stability and convergence efficiency for computations with large density and viscosity contrast.
Article
Mechanics
Jian Huang, Guanghang Wang, Yiwei Wang, Jingzhu Wang, Zhaohui Yao
Summary: This study investigates the effect of the contact angle on the generation position and focusing efficiency of annular focused jets between parallel plates. The experiment uses a pulsed laser to generate a cavitation bubble inside a droplet, leading to the formation of an annular-focused jet on the droplet surface. The study combines experiments, numerical simulations, and analytical modeling to provide insights into the flow characteristics and focusing efficiency of annular jets. The findings can contribute to high-throughput inkjet printing and liquid transfer applications.
Article
Mechanics
Yufei Wang, Zhiying Wang, Yan Du, Jingzhu Wang, Yiwei Wang, Chenguang Huang
Summary: This study investigates the dynamics of the surface-seal splash and develops an analytical model to understand its mechanics. The results indicate that the aerodynamic pressure plays a dominant role in the formation of the surface-seal splash.
Article
Mechanics
Renfang Huang, Rundi Qiu, Yuchang Zhi, Yiwei Wang
Summary: This study investigates the behaviors of ventilated cavities around a surface-piercing hydrofoil at high Froude numbers. Through experiments and modeling analysis, the study reveals the characteristics of vaporous cavitating flow and explores the transition from fully wetted flow to fully ventilated flow. The findings lay a foundation for the design optimization and control strategy of high-speed hydrofoils.
Article
Mechanics
Yao Xiao, Zhong Zeng, Liangqi Zhang, Jingzhu Wang, Yiwei Wang, Hao Liu, Chenguang Huang
Summary: In this paper, a phase-field-based spectral element method is proposed for solving the Navier-Stokes/Cahn-Hilliard equations in incompressible two-phase flows. Three constant coefficient matrixes are constructed for velocity, pressure, and phase variable solutions using the Newton-Raphson method and time-stepping scheme. The use of modified bulk free energy density ensures the boundedness of the Cahn-Hilliard equation solution. The proposed approach demonstrates high accuracy and enhanced computation efficiency for capturing interfacial dynamics.
Article
Nanoscience & Nanotechnology
Ao Wang, Yuxue Zhong, Guanghang Wang, Jian Huang, Jingzhu Wang, Yiwei Wang
Summary: The formation of two axial jets of a spark-induced bubble near a soft membrane was experimentally studied. The key parameters determining the formation of the jets were obtained, and it was found that the critical conditions for jet formation changed with the parameters. The sub-regime of membrane piercing was also discovered.
Article
Engineering, Marine
Yuchang Zhi, Rundi Qiu, Renfang Huang, Yiwei Wang
Summary: This research explores the dynamics of propeller wake under a light loading condition using dynamic mode decomposition (DMD) and reconstruction. Stable tip and hub vortices are observed without interacting evolution, and elliptical instabilities occur downstream of the tip vortices. The dominant frequencies identified by DMD are related to the blade passing frequency, and the coherent structures primarily result from the convection of the tip vortices. The reconstructed propeller wake flow shows good agreement with the original flow, indicating the potential of DMD-based flow-field reconstruction for propeller wake prediction and control.
Article
Mechanics
Yong Liu, Liangqi Zhang, Hao Liu, Linmao Yin, Yao Xiao, Yue Wang, Zhong Zeng
Summary: In this study, the instability of complex convection in the Czochralski model was explored. The results showed that the mixed convection of silicon melt becomes unstable with an increase in radii ratio, while a larger radii ratio improves the stability for LiCaAlF6 melt.
Article
Mechanics
Zhen Zhang, Jingzhu Wang, Renfang Huang, Rundi Qiu, Xuesen Chu, Shuran Ye, Yiwei Wang, Qingkuan Liu
Summary: In this study, an implicit data-driven URANS framework is proposed to analyze the unsteady characteristics of cavitating flow. A basic computational model is developed by introducing a cavitation-induced phase transition into the Reynolds stress equations. The linear and nonlinear parts of the anisotropic Reynolds stress are predicted through implicit and explicit methods, respectively, to improve the computational accuracy. The DD-URANS model is trained using numerical results obtained via large-eddy simulation and is found to outperform the baseline URANS model in predicting the unsteady characteristics of cavitating flow.
Article
Thermodynamics
Yue Wang, Liangqi Zhang, Hao Liu, Linmao Yin, Yao Xiao, Yong Liu, Zhong Zeng
Summary: The effect of Prandtl number on the instability of thermocapillary liquid bridges between coaxial disks with different radii under microgravity is investigated. It is found that the instability patterns and mechanisms depend on the radius ratio and heating strategy. The results reveal that the instability is oscillatory for small Prandtl numbers in the bottom heating scenario and the critical Marangoni number increases with the rise of Pr in the upper heating situation.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2023)
Article
Mechanics
Chang Yan, Shengfeng Xu, Zhenxu Sun, Dilong Guo, Shengjun Ju, Renfang Huang, Guowei Yang
Summary: This paper proposes a POD method, strengthened by a physics-informed neural network (PINN) with an overlapping domain decomposition strategy, to alleviate the dependence of traditional POD on the quality and quantity of data. The proposed method achieves accurate extraction of flow structures from spatially sparse observation data at different Reynolds numbers and can extract spatial structures and dominant frequency under high-level noise.