SBIR/STTR Award attributes
Illinois Rocstar LLC (IR), in collaboration with Sandia National Laboratories (SNL), will develop a robust methodology and software for automated domain decomposition and hexahedral mesh generation to address the U.S. Navy modeling and simulation (M&S) needs. The proposed automated hex meshing tool, named Auto-Hex, will be offered as an extension module to the Sandia National Laboratories CUBIT software. Auto-Hex uses advanced geometric reasoning and artificial intelligence for automated domain decomposition and robust hex(-dominant) meshing. Auto-Hex will support meshing of solid objects with user required types of input and output formats, cross-OS functionality, and compatibility with Navy M&S software. The problem of hexahedral meshing is mathematically global in scope. This geometrically constrained discretization problem is multi-objective and can be solved utilizing more than one decomposition path. Currently the most reliable way to hex mesh a complex geometry is to manually decompose it into hex-meshable partitions. This task requires significant meshing expertise, and it can take a very large portion of total M&S time. CUBIT is known for its multitude of tools for generating webcuts and domain decomposition tailored for hex-meshing. CUBIT will be enhanced with Auto-Hex functionality to form an IR in-house, cross-OS functional, automated hex(-dominant) mesh generation software. During Phase I IR and SNL teams successfully demonstrated the potential for replacing user expertise with artificial intelligence in the critical task of domain decomposition. The technologies developed during the Phase I effort included advanced geometric reasoning based on the skeletal mesh and development of machine learning models for semantic segmentation. The technical feasibility of the task was established through research, development, and prototyping of the smart decomposition and automated hex-meshing methodology. During Phase II, IR and SNL will fully develop the product. SNL will perform tasks related to skeletal-based geometric reasoning, Auto-Hex plugin GUI development, extension of the CUBIT, testing, and benchmarking. IR will perform the development, testing and verification of machine learning based decomposition, geometric loops and feature points extraction and hex meshing processes.