Table of contents

• Overview
• Contents
• Usage
• Preprocessing
• Related publications
• License
• Citation

Overview

This is the first public release of the Graph Structural Truss Dataset, intended for research in graph machine learning, topological clustering, and structural analysis. The dataset comprises weighted, undirected graphs representing networks of bars under tension and compression, in a state of static equilibrium.

The dataset was generated using a grammar-based design workflow developed by Dr. sc. Ioannis Mirtsopoulos during his doctoral research at the Structural Xploration Lab (SXL) of École polytechnique fédérale de Lausanne (EPFL), and further refined during his postdoctoral research in the Digital Structures group at the Massachusetts Institute of Technology (MIT).

The design methodology has been implemented as a plugin named Libra, developed for the parametric modeling environment Grasshopper within Rhinoceros 3D. Dataset generation employed various sampling strategies, with the outputs of each method organized into correspondingly named folders.

Contents

Each folder contains graphs generated with the sampling method indicated as the folder name. - graph.csv: per-graph properties such as static action, min/max force, and number of nodes - edges.csv: per edge properties such as start/end node id, label etc. - nodes.csv: per node properties such as x,y,z coordinates, label etc.

The structure of these .csv files is inspired by topologicpy.

Graphs.csv

| DESCRIPTION | VARIABLE NAME | | --- | --- | | The graph ID | graphid | The graph label
(0 for planar / 1 for non-planar)_ | label | The graph features:
static action [0]; min length [1]; max length [2]; mean length [3]; standard deviation length [4]; min axial force [5]; max axial force [6]; mean axial force [7]; standard deviation axial force [8] | feat

Edges.csv

| DESCRIPTION | VARIABLE NAME | | --- | --- | | The graph id it belongs to | graphid | The starting node ID | srcid | The ending node ID | dstid | The edge label
_(the edge ID) | label | The edge features:
axial force: tension < 0, compression > 0 | feat

Nodes.csv

| DESCRIPTION | VARIABLE NAME | --- | --- | | The graph ID | graphid | The node ID | nodeid | The node label
( 0 for "external" / 1 for "internal") | label | The node features:
(valency) | feat | The node X coordinate | X | The node Y coordinate | Y | The node Z coordinate | Z

Usage

Copy the .csv files locally to your system and extract the relevant information.

Preprocessing

All graphs have been cleaned and verified for connectivity, planarity, and zero force members.

Related publications

• Mirtsopoulos, I., Fivet, C., Mueller, C., 2025. “Integrating constructability constraints into an equilibrium-aware grammar for geenrative structural design”, in: Structures and Architecture. REstructure Rematerialize REthink REuse pp. 901–908. https://doi.org/10.1201/9781003658641

• Mirtsopoulos, I., Fivet, C., 2023. “Structural Topology Exploration through Policy-Based Generation of Equilibrium Representations.” Computer-Aided Design 160 (July 1, 2023): 103518. https://doi.org/10.1016/j.cad.2023.103518

• Mirtsopoulos, I., Fivet, C., 2022. "Exploration of static equilibrium representations; policies and genetic algorithms", in: Structures and Architecture A Viable Urban Perspective? pp. 1137–1144. https://doi.org/10.1201/9781003023555-136

• Mirtsopoulos, I., Fivet, C., 2021. "Grammar-based generation of bar networks in static equilibrium with bounded bar lengths", in: Proceedings of International Association for Shell and Spatial Structures (IASS) Symposium 2020/21. 23-27 August, Guilford, UK. https://doi.org/10.15126/900337

• Mirtsopoulos, I., Fivet, C., 2020. "Design space exploration through force-based grammar rule". archiDOCT 8, 50–64.

License

This dataset is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0). Please cite appropriately if used.

Citation

Please cite the DOI and credit the author.

@software{johnmirts_2025_16544554, author = {Mirtsopoulos, Ioannis}, title = {Graph Structural Truss Dataset}, month = july, year = 2025, version = {v2025-07}, doi = {10.5281/zenodo.16544554}, url = {https://doi.org/10.5281/zenodo.16544554}, }