Physics-infused deep neural network for solution of non-associative Drucker–Prager elastoplastic constitutive model

Journal of the Mechanics and Physics of Solids, Год журнала: 2024, Номер 185, С. 105570 - 105570

Опубликована: Фев. 12, 2024

Язык: Английский

---

A Comprehensive Investigation of Physics-Informed Learning in Forward and Inverse Analysis of Elastic and Elastoplastic Footing

Computers and Geotechnics, Год журнала: 2025, Номер 181, С. 107110 - 107110

Опубликована: Фев. 5, 2025

Язык: Английский

---

A conditional generative model for end-to-end stress field prediction of composite bolted joints

Engineering Applications of Artificial Intelligence, Год журнала: 2024, Номер 134, С. 108692 - 108692

Опубликована: Май 31, 2024

Язык: Английский

---

Physics-constrained deep learning approach for solving inverse problems in composite laminated plates

Composite Structures, Год журнала: 2024, Номер 348, С. 118514 - 118514

Опубликована: Авг. 23, 2024

Язык: Английский

---

Physics-informed neural network combined with characteristic-based split for solving Navier–Stokes equations

Engineering Applications of Artificial Intelligence, Год журнала: 2023, Номер 128, С. 107453 - 107453

Опубликована: Ноя. 17, 2023

Язык: Английский

---

Invariance embedded physics-infused deep neural network-based sub-grid scale models for turbulent flows

Engineering Applications of Artificial Intelligence, Год журнала: 2023, Номер 128, С. 107483 - 107483

Опубликована: Ноя. 18, 2023

In this paper, we present two novel physics-infused neural network (NN) architectures that satisfy various invariance conditions for constructing efficient and robust Deep Learning (DL)-based sub-grid scale (SGS) turbulence models use in Large Eddy Simulation (LES) procedures widely used fluid engineering applications. The first architecture, called tensor basis networks (TBNN), recently proposed the context of Reynolds-averaged Navier–Stokes (RANS) modeling, introduced herein SGS modeling wall-bounded turbulence, embeds analytical expansion stresses into integrity tensors composed symmetric anti-symmetric parts resolved velocity gradient tensor, thus incorporating Galilean, rotational reflectional invariances. our second approach, a relatively simple yet powerful Galilean Invariance embedded Neural Network (GINN) incorporates invariance, takes as inputs independent components addition to invariant single input layer. Explicit filtering data from direct simulations canonical channel flow at friction Reynolds numbers Reτ≈395 590 provided accurate training testing these models. Both sets are predict feature datasets generated with different filter widths, numbers. GINN model yields less prediction error on test (mean squared ∼0.4) compared TBNN ∼0.5). Upon comparison, it is revealed has better extraction capacity owing its ability establish relations between extract information cross-components stresses. Based their predictive performance, both have shown great promise posteriori actual LESs. work illustrates significance physics infusion well embedding NN architecture DL-based turbulent achieving superior efficacy.

Язык: Английский

---

A Self-Optimized Machine Learning Approach for Constitutive Parameters Identification of Aortic Walls

International Journal of Applied Mechanics, Год журнала: 2024, Номер 16(05)

Опубликована: Апрель 9, 2024

The accurate identification of constitutive parameters is considered the key challenge for study mechanical properties biological soft tissues. popular machine learning (ML)-based inverse frameworks always require large amounts training datasets. This work proposes an ML framework called self-optimized identifying aortic walls under few datasets conditions. includes three steps: Step 1: forward physical FEM models first simulate nonlinear deformation subject to uniaxial tension tests and are used establish relationship datasets, 2: carefully designed random forest (RF) model can offer rapid by established 3: recalled evaluate error between results in 2 real values, then accuracy embedded into RF guiding optimization directions ensure that final accurately physically reasonable. robustness validation proposed was conducted experiment samples bovine walls. approach achieves R-squared exceeding 96.90% longitudinal direction 98.30% circumferential direction, which better than directly gradient-based same respectively. comparison show not only achieve walls, but also decrease dependency initial number sampling data effectively. developed herein provides a common convenient

Язык: Английский

---

Exploring Physics-Informed Neural Networks for the Generalized Nonlinear Sine-Gordon Equation

Applied Computational Intelligence and Soft Computing, Год журнала: 2024, Номер 2024(1)

Опубликована: Янв. 1, 2024

The nonlinear sine‐Gordon equation is a prevalent feature in numerous scientific and engineering problems. In this paper, we propose machine learning‐based approach, physics‐informed neural networks (PINNs), to investigate explore the solution of generalized non‐linear equation, encompassing Dirichlet Neumann boundary conditions. To incorporate physical information for multiobjective loss function has been defined consisting residual governing partial differential (PDE), initial conditions, various Using multiple densely connected independent artificial (ANNs), called feedforward deep designed handle equations, PINNs have trained through automatic differentiation minimize that incorporates given PDE governs laws phenomena. illustrate effectiveness, validity, practical implications our proposed two computational examples from are presented. We developed PINN algorithm implemented it using Python software. Various experiments were conducted determine an optimal architecture. network training was employed by current state‐of‐the‐art optimization methods learning known as Adam L‐BFGS‐B minimization techniques. Additionally, solutions method compared with established analytical found literature. findings show approach accurate efficient solving equations variety conditions well any complex problems across disciplines.

Язык: Английский

---

PIML-SM: Physics-informed machine learning to estimate surface soil moisture from multi-sensor satellite images by leveraging swarm intelligence

IEEE Transactions on Geoscience and Remote Sensing, Год журнала: 2024, Номер 62, С. 1 - 13

Опубликована: Янв. 1, 2024

Язык: Английский

---

An improved plate deep energy method for the bending, buckling and free vibration problems of irregular Kirchhoff plates

Engineering Structures, Год журнала: 2023, Номер 301, С. 117235 - 117235

Опубликована: Дек. 14, 2023

Язык: Английский

---