SGLDBench

Code Repository: GitHub – SGLDBench
Contact: junpeng.wang@tum.de | dennis.bukenberger@tum.de

Description

This page is created to showcase and exchange the use of SGLDBench. Reporting bugs, leaving comments, and sharing cool application cases are warmly welcomed!

SGLDBench is a comprehensive benchmark suite for applying and evaluating material layout strategies to generate stiff, lightweight structural designs in 3D domains, with a special focus on high-resolution designs.

SGLDBench provides a seamlessly integrated simulation and analysis framework, including six reference strategies and a scalable multigrid elasticity solver to efficiently execute these strategies and validate the stiffness of their results.

SGLDBench is primarily implemented using MATLAB, but also has external calls to C++ code through the MEX interface for performance reasons, and to executables and Python codes for functionality reasons, which provides a handy stencil for future extensions.

For an overview of this tool, please have a look at the README of the code repository and watch the guide videos.

What can SGLDBench do?

High-resolution Topology Optimization or Porous Infill Optimization in 3D? ✔️
Design and evaluation of stiff lightweight structure in 3D using Stress-aware Graded Voronoi Diagram, Principal Stress Lines (PSLs), Stress-aligned Conforming Lattices, and Volumetric Michell Trusses? ✔️
Evaluating stiffness of external structural designs depicted in voxels, node-edge graph, tet- or hex-mesh? ✔️
Evaluating stiffness of structural designs when loading direction is changed? ✔️
Simply visualizing a structural design depicted in voxels and written in .NIFTI format? ✔️

Data Format

SGLDBench takes triangular surface meshes (.obj, .ply, .stl) as input to construct the voxelized domain for elasticity simulation.
SGLDBench introduces a tailored ASCII File (.TopVoxel) to exchange (I/O) the voxel FEA model, including solid voxels/Elements, node-wise boundary conditions, per-voxel density values (optional), and passive elements if necessary. Both element and node information are stored as the indices of the target entities in the corresponding binary volume data (V) in Row-Column-Page order, refer to indices=find(V) in MATLAB.
External lightweight structural designs depicted in a node-edge graph (.obj), tet-, or hex-mesh (.vtk, .mesh) can be imported into SGLDBench for stiffness evaluation by voxelizing graph/mesh edges
Strucctural designs made in SGLDBench are depicted in a volume data and exported in the ".NIFTI" format for downstream operations.

Related Publication

Title: "SGLDBench: A Benchmark Suite for Stress-Guided Lightweight 3D Designs"
Authors: Junpeng Wang, Dennis Bukenberger, Simon Niedermayr, Christoph Neuhauser, Jun Wu, Rüdiger Westermann
Venue: IEEE Transactions on Visualization and Computer Graphics, 2025
[Read the Paper]

Demonstration Cases

Case 3: Creating a bone-mimicking infill for a spine model via porous infill optimization

Case 2: Infill Design for a Molar

Context: Given a geometric model of a molar described in a triangular surface mesh, one can generate different types of infill patterns under the same material budget (e.g., 50%) and compare the stiffness measured by the compliance value (C).

Case 1: Topology Optimization Practices with SGLDBench

Contexts: With SGLDBench, beginners stepping into high-resolution topology optimization can quickly create topology-optimized high-resolution designs, exploring the effects of different boundary conditions, simulation resolutions, filtering radius. SGLDBench also integrates several simple shapes (cuboid, L-shape, cylinder) to facilitate this.

Practice 1.1 Same cuboid design domain but with different boundary conditions. Resolution: 300x150x150