arXiv subjects

Jidong Zhao

Publications and source records attributed to Jidong Zhao.

3 recordsLinked to original sources

arXiv2026

An elasto-viscoplastic thixotropic model for fresh concrete capturing flow-rest transition

The flow properties of fresh concrete are critical in the construction industry, as they directly affect casting quality and the durability of the final structure. Although non-Newtonian fluid models, such as the Bingham model, are widely used to model these flow properties, they often fail to capture key phenomena, including flow stoppage, and frequently rely on ad hoc, non-physical regularization or stabilization techniques to mitigate numerical instabilities at low shear rates. To address these limitations, this study proposes an elasto-viscoplastic constitutive model within the continuum mechanics framework, which treats fresh concrete as a solid-like material with a rate-dependent yield stress. The model inherently captures the transition from elastic response to viscous flow following Bingham rheology, and vice versa, enabling accurate prediction of flow cessation without ad hoc criteria. Additionally, a thixotropy evolution law is incorporated to account for the time-dependent behavior resulting from physical flocculation and shear-induced deflocculation. The proposed model is implemented within the Material Point Method (MPM), whose Lagrangian formulation facilitates tracking of history-dependent variables and robust simulation of large deformation flows. Numerical examples demonstrate the model's effectiveness in reproducing a range of typical concrete flow scenarios, offering a more physically consistent numerical tool for optimizing concrete construction processes and minimizing defects.

cond-mat.soft↗

arXiv2026

FEM-Informed Hypergraph Neural Networks for Efficient Elastoplasticity

Graph neural networks (GNNs) naturally align with sparse operators and unstructured discretizations, making them a promising paradigm for physics-informed machine learning in computational mechanics. Motivated by discrete physics losses and Hierarchical Deep Learning Neural Network (HiDeNN) constructions, we embed finite-element (FEM) computations at nodes and Gauss points directly into message-passing layers and propose a numerically consistent FEM-Informed Hypergraph Neural Networks (FHGNN). Similar to conventional physics-informed neural networks (PINNs), training is purely physics-driven and requires no labeled data: the input is a node element hypergraph whose edges encode mesh connectivity. Guided by empirical results and condition-number analysis, we adopt an efficient variational loss. Validated on 3D benchmarks, including cyclic loading with isotropic/kinematic hardening, the proposed method delivers substantially improved accuracy and efficiency over recent, competitive PINN variants. By leveraging GPU-parallel tensor operations and the discrete representation, it scales effectively to large elastoplastic problems and can be competitive with, or faster than, multi-core FEM implementations at comparable accuracy. This work establishes a foundation for scalable, physics-embedded learning in nonlinear solid mechanics.

cs.LG↗

arXiv2026

A Virtual Heat Flux Method for Simple and Accurate Neumann Thermal Boundary Imposition in the Material Point Method

In the Material Point Method (MPM), accurately imposing Neumann-type thermal boundary conditions, particularly convective heat flux boundaries, remains a significant challenge due to the inherent nonconformity between complex evolving material boundaries and the fixed background grid. This paper introduces a novel Virtual Heat Flux Method (VHFM) to overcome this limitation. The core idea is to construct a virtual flux field on an auxiliary domain surrounding the physical boundary, which exactly satisfies the prescribed boundary condition. This transforms the surface integral in the weak form into an equivalent, and easily computed, volumetric integral. Consequently, VHFM eliminates the need for explicit boundary tracking, specialized boundary particles, or complex surface reconstruction. A unified formulation is presented, demonstrating the method's straightforward extension to general scalar, vector, and tensor Neumann conditions. The accuracy, robustness, and convergence of VHFM are rigorously validated through a series of numerical benchmarks, including 1D transient analysis, 2D and 3D curved boundaries, and problems with large rotations and complex moving geometries. The results show that VHFM achieves accuracy comparable to conforming node-based imposition and significantly outperforms conventional particle-based approaches. Its simplicity, computational efficiency, and robustness make it an attractive solution for integrating accurate thermal boundary conditions into thermo-mechanical and other multiphysics MPM frameworks.

math-ph↗