Conventional density-based topology optimization often produces optimized structures with indistinct or stair-stepped boundaries, making it difficult to convert the results into practical design models and requiring significantly higher computational ...
Conventional density-based topology optimization often produces optimized structures with indistinct or stair-stepped boundaries, making it difficult to convert the results into practical design models and requiring significantly higher computational cost when high-resolution meshes are employed. To overcome these limitations, the present work introduces a density-based topology optimization framework where nodal movements, alongside density variables, are incorporated into the set of design variables, enabling deformation of the reference mesh. This approach enhances geometric resolution without the need for highly refined discretization and removes the need for additional post-processing procedures by yielding a binary density field with well-defined smooth boundaries.
The proposed framework is primarily developed on unstructured triangular meshes, enabling flexible representation of complex geometries. A patch-based approach is employed to suppress non-physical features, such as one-node connections, while promoting density field binarization. To ensure element validity and numerical stability under nodal movements, geometric constraints based on element shape information are incorporated, and an enriched finite element method is employed to maintain analysis accuracy even when the mesh becomes distorted. By integrating these formulations, a unified framework is developed that maintains mesh quality throughout the optimization procedure. The effectiveness of this framework is verified across diverse problems with complex geometries.
The proposed approach is also implemented on structured quadrilateral meshes using standard finite elements for direct comparison with conventional high-resolution topology optimization. In this formulation, additional functions are incorporated to preserve mesh quality and maintain smooth boundaries during optimization. Through this comparison, it is confirmed that the proposed framework achieves efficient topology optimization using fewer elements while reducing computational time.