Virtual topology based hex-dominant meshing and re-meshing

Abstract

Computer-aided engineering methods have been increasingly used in various industries to support the design of complex systems and validate their performance by simulating their physical behaviour. This created a need to generate fit-for-purpose analysis models with a turnaround time comparable to the time required to generate the design model. However, most existing tools for computer-aided design and engineering have been developed independently to address different issues and therefore lack integration. As a consequence, many current simulation workflows rely on experienced users to adapt, or create from scratch, an appropriate analysis model to represent the physics to be solved. Preparing complex multi-physics analysis models for very large assemblies can take up to several person-years, limiting the tools for validation rather than having the maximum impact on design decisions.

This thesis builds on previous research on automatic hex-mesh generation and CAD-CAE integration to develop a novel framework for generating CAE analysis models from CAD design models. The framework uses ‘virtual topology’ to bridge the representation gap between CAD and CAE tools, overcome geometric limitations and accommodate the multiple different representations necessary for analysis workflows. Virtual topology is also a key enabler for the flexible modification and updating of the analysis models. Robust information management is achieved by implementing a cellular model within the virtual topology framework, which hosts and interconnects the different representations while capturing the information necessary for downstream processes.

In this work, 9 pre-processing reasoners are defined within the virtual topology framework to automatically generate hex-dominant meshes by incrementally identifying and extracting virtual regions for which a simple hex-meshing strategy can be applied. A new translator tool is introduced to query virtual models using virtual geometry curves. Additional reasoners are then used to automatically process the different meshing strategies and generate a conformal mesh, reducing the CAD to mesh time from several hours to several minutes compared to manual decomposition and existing meshing methods. Incorporating meshing constraints early in the pre-processing workflow also enables the enhancement of the pre-processing tools. A new method to propagate design modifications is developed, so that a decomposed analysis model and its associated mesh can be updated after topological changes on the design model. This is up to 95% faster than re-starting the automated pre-processing from the beginning.

Finally, examples of aeronautical components are processed using the virtual topology framework developed to demonstrate the benefits of virtual topology based workflows for hex-dominant mesh generation.

Date of Award	Dec 2020
Original language	English
Awarding Institution	Queen's University Belfast
Sponsors	Rolls Royce PLC
Supervisor	Trevor T Robinson (Supervisor), Cecil Armstrong (Supervisor) & Christopher Tierney (Supervisor)