Introducing a microstructure-embedded autoencoder approach for reconstructing high-resolution solution field data from a reduced parametric space DOI Creative Commons
Rasoul Najafi Koopas, Shahed Rezaei, Natalie Rauter

et al.

Computational Mechanics, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 21, 2024

Abstract In this study, we develop a novel multi-fidelity deep learning approach that transforms low-fidelity solution maps into high-fidelity ones by incorporating parametric space information an autoencoder architecture. This method’s integration of significantly reduces the amount training data needed to effectively predict solutions from ones. examine two-dimensional steady-state heat transfer analysis within heterogeneous materials microstructure. The conductivity coefficients for two different are condensed 101 $$\times $$ × grid smaller grids. We then solve boundary value problem on coarsest using pre-trained physics-informed neural operator network known as Finite Operator Learning (FOL). resulting is subsequently upscaled back newly designed enhanced autoencoder. novelty developed lies in concatenation resolutions decoder segment distinct steps. Hence algorithm named microstructure-embedded (MEA). compare MEA outcomes with those finite element methods, standard U-Net, and interpolation upscaling technique. Our shows outperforms these methods terms computational efficiency error representative test cases. As result, serves potential supplement networks, while preserving critical details often lost traditional such sharp interfaces features context approaches.

Language: Английский

Pre-trained Physics-Informed Neural Networks for Analysis of Contaminant Transport in Soils DOI

Zhenyu Ke,

Sheng-Jie Wei,

Shi-Yuan Yao

et al.

Computers and Geotechnics, Journal Year: 2025, Volume and Issue: 180, P. 107055 - 107055

Published: Jan. 13, 2025

Language: Английский

Citations

0
Physics-Informed Neural Network-Based Discovery of Hyperelastic Constitutive Models from Extremely Scarce Data DOI

Hyun Su Moon,

Donggeun Park,

Hanbin Cho

et al.

Published: Jan. 1, 2025

Language: Английский

Citations

0
Piecewise physics-informed neural networks for surrogate modelling of non-smooth system in elasticity problems using domain decomposition DOI
Youngjoon Jeong, Sang-ik Lee, Jong-hyuk Lee

et al.

Biosystems Engineering, Journal Year: 2025, Volume and Issue: 251, P. 48 - 60

Published: Feb. 7, 2025

Language: Английский

Citations

0
Preconditioned FEM-based neural networks for solving incompressible fluid flows and related inverse problems DOI
Franziska Griese, Fabian Hoppe, Alexander Rüttgers

et al.

Journal of Computational and Applied Mathematics, Journal Year: 2025, Volume and Issue: unknown, P. 116663 - 116663

Published: April 1, 2025

Language: Английский

Citations

0
Deep learning-driven medical image analysis for computational material science applications DOI Creative Commons
Lu Li,

Mingpei Liang

Frontiers in Materials, Journal Year: 2025, Volume and Issue: 12

Published: April 8, 2025

Introduction Deep learning has significantly advanced medical image analysis, enabling precise feature extraction and pattern recognition. However, its application in computational material science remains underexplored, despite the increasing need for automated microstructure analysis defect detection. Traditional processing methods often rely on handcrafted threshold-based segmentation, which lack adaptability to complex microstructural variations. Conventional machine approaches struggle with data heterogeneity extensive labeled datasets. Methods To overcome these limitations, we propose a deep learning-driven framework that integrates convolutional neural networks (CNNs) transformer-based architectures enhanced representation. Our method incorporates domain-adaptive transfer multi-modal fusion techniques improve generalizability of analysis. Results Experimental evaluations diverse datasets demonstrate superior performance segmentation accuracy, detection robustness, efficiency compared traditional methods. Discussion By bridging gap between science, our approach contributes more effective, automated, scalable characterization processes.

Language: Английский

Citations

0
A Finite Operator Learning Technique for Mapping the Elastic Properties of Microstructures to Their Mechanical Deformations DOI Creative Commons
Shahed Rezaei, Reza Najian Asl,

Shirko Faroughi

et al.

International Journal for Numerical Methods in Engineering, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 9, 2024

ABSTRACT To obtain fast solutions for governing physical equations in solid mechanics, we introduce a method that integrates the core ideas of finite element with physics‐informed neural networks and concept operators. We propose directly utilizing available discretized weak form packages to construct loss functions algebraically, thereby demonstrating ability find even presence sharp discontinuities. Our focus is on micromechanics as an example, where knowledge deformation stress fields given heterogeneous microstructure crucial further design applications. The primary parameter under investigation Young's modulus distribution within system. investigations reveal physics‐based training yields higher accuracy compared purely data‐driven approaches unseen microstructures. Additionally, offer two methods improve process obtaining high‐resolution solutions, avoiding need use basic interpolation techniques. first one based autoencoder approach enhance efficiency calculation high resolution grid points. Next, Fourier‐based parametrization utilized address complex 2D 3D problems micromechanics. latter idea aims represent microstructures efficiently using Fourier coefficients. proposed draws from deep energy but generalizes enhances them by learning parametric without relying external data. Compared other operator frameworks, it leverages domain decomposition several ways: (1) uses shape derivatives instead automatic differentiation; (2) automatically includes node connectivity, making solver flexible approximating jumps solution fields; (3) can handle arbitrary shapes enforce boundary conditions. provided some initial comparisons well‐known algorithms, emphasize advantages newly method.

Language: Английский

Citations

2
Prediction of microstructural evolution of multicomponent polymers by Physics-Informed neural networks DOI

Jiaqi An,

Yanlong Ran,

Jiaping Lin

et al.

Computational Materials Science, Journal Year: 2024, Volume and Issue: 246, P. 113502 - 113502

Published: Nov. 4, 2024

Language: Английский

Citations

0
Introducing a microstructure-embedded autoencoder approach for reconstructing high-resolution solution field data from a reduced parametric space DOI Creative Commons
Rasoul Najafi Koopas, Shahed Rezaei, Natalie Rauter

et al.

Computational Mechanics, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 21, 2024

Abstract In this study, we develop a novel multi-fidelity deep learning approach that transforms low-fidelity solution maps into high-fidelity ones by incorporating parametric space information an autoencoder architecture. This method’s integration of significantly reduces the amount training data needed to effectively predict solutions from ones. examine two-dimensional steady-state heat transfer analysis within heterogeneous materials microstructure. The conductivity coefficients for two different are condensed 101 $$\times $$ × grid smaller grids. We then solve boundary value problem on coarsest using pre-trained physics-informed neural operator network known as Finite Operator Learning (FOL). resulting is subsequently upscaled back newly designed enhanced autoencoder. novelty developed lies in concatenation resolutions decoder segment distinct steps. Hence algorithm named microstructure-embedded (MEA). compare MEA outcomes with those finite element methods, standard U-Net, and interpolation upscaling technique. Our shows outperforms these methods terms computational efficiency error representative test cases. As result, serves potential supplement networks, while preserving critical details often lost traditional such sharp interfaces features context approaches.

Language: Английский

Citations

0