2024年1月10日 · To address these issues, we propose structure-preserving PINNs to improve their performance and broaden their applications for downstream tasks. Firstly, by leveraging prior knowledge about the physical system, a structure-preserving loss function is designed to assist the PINN in learning the underlying structure.

2024年3月1日 · A PINN approach can map the spatiotemporal coordinates to the solution of the PDEs, and the output w ˜ (x, t, θ) is regarded as the approximate solution of w (x, t). The core algorithm of the PINN approach was first proposed by Raissi and his co-workers [10], and its working principle is shown in Fig. 2.

2023年1月1日 · Physics-Informed Neural Network (PINN) predicts the small-strain response of curved shells. Mesh-free solution of the underlying Naghdi shell equations in non-Euclidean domains. Scordelis–Lo roof test confirms accurate solutions even in the small-thickness limit.

2023年12月1日 · Inspired by concepts of PINNs, this paper proposes a novel framework, ‘Complete Physics-Informed Neural Network-based Topology Optimization (CPINNTO)’, to address various challenges in topology optimization, particularly related to structural optimization.

2022年5月27日 · Here, we propose Auto-PINN, the first systematic, automated hyperparameter optimization approach for PINNs, which employs Neural Architecture Search (NAS) techniques to PINN design. Auto-PINN avoids manually or exhaustively searching the hyperparameter space associated with PINNs.

To address these problems, we propose a new family of PINNs, named structure-preserving PINNs (SP-PINNs), that can be applied to dynamical systems with energy or Lya-punov structure. Our approach is capable of solving PDEs (e.g., the Allen–Cahn equation) and handling downstream tasks (e.g., image recognition).

2022年7月26日 · As a potential alternative, we use machine learning and present a Physics-Informed Neural Network (PINN) to predict the small-strain response of arbitrarily curved shells. To this end, the shell midsurface is described by a chart, from which the mechanical fields are derived in a curvilinear coordinate frame by adopting Naghdi's shell theory.

物理信息神经网络 (PINN) 在深度学习模型的训练中包含支配现实的物理定律，从而能够对复杂现象进行预测和建模，同时遵守基本物理原理。 PINN 是一类物理信息 机器学习 方法，可将物理知识与数据无缝集成。 在求解涉及 PDE 和 ODE 的问题时，通常会将 PINN 与纯数据驱动方法和传统数值方法进行比较。 纯粹数据驱动方法仅从输入和输出数据中学习数学关系，而 PINN 与之不同： 使用先验物理知识。 在训练数据集之外作出更准确的预测。 在训练数据有限或含噪的情况下 …

3 天之前 · Applications in physics: (a) Input 2D wide-angle scattering pattern, the reconstructed & ground truth nanoscale structure and the simulated scattering pattern using the predicted structure . (b) A schematic from [ 303 ], where a PINN is used to track the trajectory of space debris after inelastic collision with a satellite.

Physics-informed neural networks (PINNs), [1] also referred to as Theory-Trained Neural Networks (TTNs), [2] are a type of universal function approximators that can embed the knowledge of any physical laws that govern a given data-set in the learning process, and can be described by partial differential equations (PDEs).