Graph Rewriting for Structural Connection Design: A Semi-Automated Extraction-to-Generation Framework

Ho Man Yau; Davide Lombardi; Theodoros  Dounas; Asterios Agkathidis; Yang Song

Graph Rewriting for Structural Connection Design: A Semi-Automated Extraction-to-Generation Framework

Ho Man Yau
, Davide Lombardi^*
, Theodoros Dounas
, Asterios Agkathidis
, Yang Song

^*Corresponding author for this work

Heriot-Watt University
University of Liverpool

Research output: Chapter in Book or Report/Conference proceeding › Conference Proceeding › peer-review

Abstract

Graph-based representations provide a powerful relational framework for modelling and analysing complex systems and enabling generative exploration. This paper proposes and validates a semi-automated graph extraction-to-generation framework prototype that bridges between 3D geometric modelling and rule-based graph rewriting. The prototype is implemented in Python using TopologicPy for semi-automated extraction and semantic classification. The resulting graph dataset enables rule mining and graph rewriting through a structured grammar system. Experiments conducted on a library of concave and convex joint components demonstrate the feasibility of scalable graph preparation and controlled generative transformation. The framework provides a reproducible workflow for transforming geometric data into generative-ready graph structures and lays the foundation for future integration with graph neural networks and reinforcement-learning-based design optimisation.

Original language	English
Title of host publication	Informed creativity and fabrication in architecture and engineering
Subtitle of host publication	Proceedings of the 44th Annual Conference of the Association of Education and Research in Computer Aided Architectural Design in Europe
Publication status	Accepted/In press - May 2026

Cite this

Yau, H. M., Lombardi, D., Dounas, T., Agkathidis, A., & Song, Y. (Accepted/In press). Graph Rewriting for Structural Connection Design: A Semi-Automated Extraction-to-Generation Framework. In Informed creativity and fabrication in architecture and engineering: Proceedings of the 44th Annual Conference of the Association of Education and Research in Computer Aided Architectural Design in Europe

Yau, Ho Man ; Lombardi, Davide ; Dounas, Theodoros et al. / Graph Rewriting for Structural Connection Design : A Semi-Automated Extraction-to-Generation Framework. Informed creativity and fabrication in architecture and engineering: Proceedings of the 44th Annual Conference of the Association of Education and Research in Computer Aided Architectural Design in Europe. 2026.

@inproceedings{277d0449c7734645bd8bf11825184ad3,

title = "Graph Rewriting for Structural Connection Design: A Semi-Automated Extraction-to-Generation Framework",

abstract = "Graph-based representations provide a powerful relational framework for modelling and analysing complex systems and enabling generative exploration. This paper proposes and validates a semi-automated graph extraction-to-generation framework prototype that bridges between 3D geometric modelling and rule-based graph rewriting. The prototype is implemented in Python using TopologicPy for semi-automated extraction and semantic classification. The resulting graph dataset enables rule mining and graph rewriting through a structured grammar system. Experiments conducted on a library of concave and convex joint components demonstrate the feasibility of scalable graph preparation and controlled generative transformation. The framework provides a reproducible workflow for transforming geometric data into generative-ready graph structures and lays the foundation for future integration with graph neural networks and reinforcement-learning-based design optimisation.",

author = "Yau, \{Ho Man\} and Davide Lombardi and Theodoros Dounas and Asterios Agkathidis and Yang Song",

year = "2026",

month = may,

language = "English",

booktitle = "Informed creativity and fabrication in architecture and engineering",

}

Yau, HM , Lombardi, D, Dounas, T, Agkathidis, A & Song, Y 2026, Graph Rewriting for Structural Connection Design: A Semi-Automated Extraction-to-Generation Framework. in Informed creativity and fabrication in architecture and engineering: Proceedings of the 44th Annual Conference of the Association of Education and Research in Computer Aided Architectural Design in Europe.

Graph Rewriting for Structural Connection Design: A Semi-Automated Extraction-to-Generation Framework. / Yau, Ho Man ; Lombardi, Davide; Dounas, Theodoros et al.
Informed creativity and fabrication in architecture and engineering: Proceedings of the 44th Annual Conference of the Association of Education and Research in Computer Aided Architectural Design in Europe. 2026.

Research output: Chapter in Book or Report/Conference proceeding › Conference Proceeding › peer-review

TY - GEN

T1 - Graph Rewriting for Structural Connection Design

T2 - A Semi-Automated Extraction-to-Generation Framework

AU - Yau, Ho Man

AU - Lombardi, Davide

AU - Dounas, Theodoros

AU - Agkathidis, Asterios

AU - Song, Yang

PY - 2026/5

Y1 - 2026/5

N2 - Graph-based representations provide a powerful relational framework for modelling and analysing complex systems and enabling generative exploration. This paper proposes and validates a semi-automated graph extraction-to-generation framework prototype that bridges between 3D geometric modelling and rule-based graph rewriting. The prototype is implemented in Python using TopologicPy for semi-automated extraction and semantic classification. The resulting graph dataset enables rule mining and graph rewriting through a structured grammar system. Experiments conducted on a library of concave and convex joint components demonstrate the feasibility of scalable graph preparation and controlled generative transformation. The framework provides a reproducible workflow for transforming geometric data into generative-ready graph structures and lays the foundation for future integration with graph neural networks and reinforcement-learning-based design optimisation.

AB - Graph-based representations provide a powerful relational framework for modelling and analysing complex systems and enabling generative exploration. This paper proposes and validates a semi-automated graph extraction-to-generation framework prototype that bridges between 3D geometric modelling and rule-based graph rewriting. The prototype is implemented in Python using TopologicPy for semi-automated extraction and semantic classification. The resulting graph dataset enables rule mining and graph rewriting through a structured grammar system. Experiments conducted on a library of concave and convex joint components demonstrate the feasibility of scalable graph preparation and controlled generative transformation. The framework provides a reproducible workflow for transforming geometric data into generative-ready graph structures and lays the foundation for future integration with graph neural networks and reinforcement-learning-based design optimisation.

M3 - Conference Proceeding

BT - Informed creativity and fabrication in architecture and engineering

ER -