An adaptive hexahedral mesh generation method for industrial CT images

Abstract

This article presents a method for the direct generation of an adaptive hexahedral mesh from industrial computed tomography (CT) images. First, the two-dimensional (2D) Shi-Tomasi algorithm is extended to the three-dimensional (3D) space to extract the 3D curvature feature points of the workpiece, and the initial non-uniform core mesh is obtained by adaptive partitioning of the octree based on the feature points information. Subsequently, to address the issue of hanging nodes within the initial non-uniform core mesh, a search template for hanging node transition chain is devised to locate these nodes, and the template method is employed to eliminate the hanging nodes from the transition chain. Then, to enhance the quality of the boundary mesh, the vector projection method is introduced for surface fitting of the core mesh, and the Laplace algorithm of vector projection optimization is combined to optimize the boundary mesh. Finally, the adaptive hexahedral mesh is generated directly from the industrial CT images obtained through actual scanning. Quality evaluation results demonstrate that the scaled Jacobian of the hexahedral mesh exceeds 0, which confirms the correctness and effectiveness of the proposed method.