A nonlocal energy‐informed neural network for isotropic elastic solids with cracks under thermomechanical loads

International Journal for Numerical Methods in Engineering, Journal Year: 2023, Volume and Issue: 124(18), P. 3935 - 3963

Published: May 25, 2023

Abstract In this paper, a nonlocal energy‐informed neural network is proposed to characterize the deformation behaviors of elastic solids with cracks subjected thermomechanical loads in framework ordinary state‐based peridynamics. Based on principle virtual work, energy approach developed recast solution peridynamic equilibrium equation as problem minimizing potential system, which automatically satisfies traction‐free boundary conditions. Meanwhile, representation physical system can be treated loss function for machine learning methods. Therefore, constructed approximate system. A distinct advantage that strain expressed terms spatial integration instead derivatives, avoids invalidation automatic differentiation at crack surfaces original physics‐informed networks. To demonstrate convergence and accuracy network, series problems solid fracture mechanics are conducted, compared results from analytical solutions or classical numerical Additionally, material containing initial cracks, displacement extrapolation method encoded into evaluate static stress intensity factor.

Language: Английский

---

Comparative assessment for pressure field reconstruction based on physics-informed neural network

Physics of Fluids, Journal Year: 2023, Volume and Issue: 35(7)

Published: July 1, 2023

In this paper, a physics-informed neural network (PINN) is used to determine pressure fields from the experimentally measured velocity data. As novel method of data assimilation, PINN can simultaneously optimize and solve by embedding Navier–Stokes equations into loss function. The compared with two traditional reconstruction algorithms, i.e., spectral decomposition-based fast integration irrotation correction on gradient orthogonal-path integration, its performance numerically assessed using kinds flow motions, namely, Taylor's decaying vortices forced isotropic turbulence. case two-dimensional vortices, critical parameters have been investigated without considering measurement errors. Regarding turbulence, influence spatial resolution out-of-plane motion assessed. Finally, in an experimental synthetic jet impinging solid wall, obtained planar particle image velocimetry. All results show that PINN-based superior other methods even if are significantly contaminated

Language: Английский

---

Pre-Training Physics-Informed Neural Network with Mixed Sampling and Its Application in High-Dimensional Systems

Journal of Systems Science and Complexity, Journal Year: 2024, Volume and Issue: 37(2), P. 494 - 510

Published: Jan. 26, 2024

Language: Английский

---

Quantification of gradient energy coefficients using physics-informed neural networks

International Journal of Mechanical Sciences, Journal Year: 2024, Volume and Issue: 273, P. 109210 - 109210

Published: March 25, 2024

Language: Английский

---

Transfer learning enhanced nonlocal energy-informed neural network for quasi-static fracture in rock-like materials

Computer Methods in Applied Mechanics and Engineering, Journal Year: 2024, Volume and Issue: 430, P. 117226 - 117226

Published: July 25, 2024

Language: Английский

---