MeshLink: A surface structured mesh generation framework to facilitate automated data linkage

Abstract

With the rapid progress in engineering simulation, there is an increasing industrial need for accurate mesh generation methods. Furthermore, the development of data-driven methods require an innovative mesh generation framework that can integrate deep learning models to facilitate automatic data linkage. This paper develops a surface structured mesh generation framework named MeshLink. This framework extends the support for mesh data and associated algorithms through the Mesh-based Feature Information Framework (MFIF). First, in order to map the model to the parametric domain, we discretize the input model using triangular meshes. To handle complex geometric shapes, we develop a structured mesh generation technique based on conformal mapping. Then, we generate the surface structured mesh of the model based on inverse mapping algorithm. The MeshLink framework overcomes the limitations of traditional mesh generation workflows by integrating deep learning models. We adopted a structured mesh evaluation model based on graph neural network, which enhance the efficiency of the framework. Finally, based on the mesh quality evaluation results, we use the corresponding mesh optimization algorithm to generate high-quality surface structure meshes. The MeshLink framework not only provides a tool that supports high-quality surface structured mesh generation, but also facilitates the storage, linking and retrieval of mesh data sources.