CJChuanming Ju

Chuanming Ju

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Chuanming Ju has a total of 30 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • BEPGN: A boundary element-enhanced physics-informed graph neural network for solving Laplace boundary value problems

    by Muwei Wang, Chuanming Ju, Yifei Wang, Huaihou Zhu, Zhuolin Liu, Baotao Chi, Guoqing Ma, Yantao Wang

  • Open Access

    A Novel Edge-Based Graph Convolutional Network Algorithm for Mesh Smoothing

    by Chuanming Ju, Jiehao Chen, Muwei Wang, Huaihou Zhu, Yifei Wang, Baotao Chi, Guoqing Ma, Yantao Wang

    This paper proposes an edge-based graph convolutional neural network (EGCN-Smooth) mesh smoothing algorithm, aiming to enhance the quality and optimization efficiency of unstructured meshes. Traditional optimization-based smoothing methods can generate high-quality meshes but incur high computational costs, while existing neural network approaches that directly process node coordinate sequences struggle to capture the topological structure. of local meshes and exhibit limited generalization capabilities. To address these issues, this paper adopts the free edge as the fundamental optimization unit, constructs a training dataset oriented toward edge optimization, and integrates graph convolutional neural networks to uniformly encode the geometric and topological features of local meshes. The EGCN-Smooth model employs an end-to-end processing approach, directly predicting the optimized coordinates of free edge endpoints and completing the smoothing operation in a single forward pass, significantly reducing iterative computational overhead. Experiments demonstrate that on various two-dimensional planar meshes and three-dimensional surface meshes, the proposed method outperforms traditional smoothing algorithms across multiple metrics, including minimum angle, average minimum angle, and element shape quality, providing a scalable neural network solution for efficient and high-quality mesh smoothing.

  • Open Access

    Graph Neural Network-Based Smoothing Method for 3D Hexahedral Meshes

    by Chuanming Ju, Huaihou Zhu, Yifei Wang, Muwei Wang, Jiehao Chen, Zhuolin Liu

    Mesh quality has a critical impact on the accuracy, convergence, and computational efficiency of numerical simulations. This paper addresses the smoothing problem of three-dimensional hexahedral structured meshes and proposes a mesh smoothing method based on Graph Neural Networks, abbreviated as GNN smoothing. The method formulates mesh smoothing as a regression problem of nodal displacement prediction: it takes interior free nodes and their local subgraphs as input, uses displacements derived from Jacobian-based optimization as supervised labels, and trains a GNN model to directly predict the optimal nodal movements. Employing the Jacobian ratio as the mesh quality metric, numerical experiments are conducted on four representative three-dimensional hexahedral mesh models: a flange, a R-type part, a shaft, and a wrench. The results demonstrate that the proposed GNN smoothing method can effectively improve the initial mesh quality, achieving mesh quality comparable to that of traditional optimization-based methods while offering significantly faster computational speed, which validates the effectiveness and practicality of the proposed method.

  • Open Access

    BEPGN: A Boundary Element-Enhanced Physics-Informed Graph Neural Network for Solving Laplace Boundary Value Problems

    by Muwei Wang, Chuanming Ju, Yifei Wang, Huaihou Zhu, Zhuolin Liu, Baotao Chi, Guoqing Ma, Yantao Wang

    This paper proposes a Boundary Element-Enhanced Physics-Informed Graph Neural Network method (BEPGN), which integrates the Boundary Element Method, Physics-Informed Neural Networks, and Graph Neural Networks. Unlike existing BEM-based neural network solvers, which rely on feedforward networks and treat boundary elements independently, BEPGN constructs a graph from the discretized boundary and employs message passing to explicitly capture the topological relationships between adjacent elements, with the boundary integral equation residual serving directly as the physics-informed loss. Numerical experiments under Dirichlet, Neumann, and mixed boundary conditions demonstrate that BEPGN achieves accuracy comparable to traditional BEM and converges faster than the Boundary Element Method based Physics-Informed Neural Network approach (BEM-PINN), validating the advantage of graph-structured learning in boundary element-based physics-informed solvers.

  • Automatic mesh generation for electromagnetic simulation on complex industry products

    by Shaojing Li, Chuanming Ju, JianFei Liu

    Purpose: This paper aims to develop a novel scheme for automatically generating high-quality tetrahedral meshes for electromagnetic simulations on complex industrial products, even in the presence of geometric defects.Design/methodology/approach: This paper substantially revises and improves three techniques namely cut-and-fix, sphere-packing and small polyhedron reconnection (SPR) and integrates them into a novel scheme. The scheme involves three key steps: first, repairing models with geometric defects using an enhanced cut-and-fix technique; second, applying sphere-packing to place nodes that capture the feature sizes of the model and finally, performing a standard Delaunay triangulation of these nodes, followed by boundary recovery and mesh quality improvement using SPR.Findings: The revised cut-and-fix method focuses on surface smoothness and effectively avoids the rough surfaces generated during the global tetrahedron optimization process. The revised sphere-packing method has the ability for meshing size adaptation and is suitable for electromagnetic simulations. It also helps to integrate cut-and-fix, sphere-packing and SPR skillfully into a novel meshing scheme.Originality/value: Numerical examples demonstrate the effectiveness and robustness of the proposed method for meshing various types of electromagnetic simulation models, highlighting its excellent performance.

  • A binary-tree subdivision method for evaluation of singular integrals with discontinuous kernel in 3D BEM

    by Chuanming Ju, Jiehao Chen, Ning Li, Xianfeng Du

    Purpose: A binary-tree subdivision method (BTSM) for numerical evaluation of weakly singular integrals with discontinuous kernel in the three-dimensional (3D) boundary element method (BEM) is presented in this paper.Design/methodology/approach: In this method, the singular boundary element is split into two sub-elements and subdivided recursively until the termination criterion is met and the subdivision is stopped. Then, the source point is surrounded by one or more spherical cavities determined by the discontinuous kernel function. The sub-elements located in spherical cavities will be eliminated, and the regular triangular or rectangle elements are employed to fill the spherical cavities.Findings: With the proposed method, the obtained sub-elements are automatically refined as they approach the source point, and they are “good” in shape and size for standard Gaussian quadrature. Thus, the proposed method can be used to evaluate singular integrals owing discontinuous kernel function accurately for cases of different element shapes and various source point locations.Originality/value: Numerical examples show that the BTSM is suitable for planar and curved elements of arbitrary regular or irregular shape at various source point locations, and the results have much better accuracy and robustness than conventional subdivision method (CSM) when the kernel function is discontinuous.


  • An adaptive binary-tree subdivision method for evaluation of nearly singular integrals in 3D BEM

    by Chuanming Ju, J. Zhang, Yudong Zhong, Xianfeng Du, Jun Li, Baotao Chi

    Purpose: The purpose of this paper is to present an adaptive binary-tree element subdivision method (BTSM) for the evaluation of nearly singular integrals in three-dimensional boundary element method, which can facilitate automatic and high-quality patch generation. Design/methodology/approach: In this method, the nearly singular element is split into two sub-elements. Each sub-element is then examined to determine if it is to be subdivided based on a specific subdivision criterion. The specific subdivision ensures that those sub-elements far from the source point are sparse. And then those sub-elements in close proximity to the source point are replaced by regular triangular elements. Findings: With the proposed method, the sub-elements obtained are automatically refined as they approach the projection point, and they are “good” in shape and size for standard Gaussian quadrature. Thus, the proposed method can be used to evaluate nearly singular integrals accurately for cases of different element shapes and various locations of the source point. Originality/value: Numerical examples for surface elements with various relative locations of the source point are presented. The results demonstrate that the proposed method has much better accuracy and robustness than some other methods.