Machine Learning For Funicular Floor Design


This research proposes a geometry-based generative design method that utilizes Machine Learning to generate various funicular floor structures with simple sketch input. The force subdivision method has been used in generating compression-dominant floor structures. In addition, machine learning techniques such as artificial neural networks have been used to evaluate and optimize the reciprocal structural forms after subdividing the force diagram.  Subsequently, a computational algorithm has been developed to identify the relationship between columns and walls in a floor plan and generate a force diagram for the construction of a network of funicular systems between the walls and columns. Subsequently, the algorithm was combined with the machine learning technique to automatically generate the structure from a sketch input of columns and walls on a plan drawing. This interdisciplinary approach combines computer science, and structural and architectural design to provide flexible design solutions for building structures.

Principal Investigators: Masoud Akbarzadeh
Lead Researcher: Hao Zheng
Research Team: Hao Zheng, Xinyu Wang, Zehua Qi, Shixuan Sun 
Archietctural Design Team: Masoud Akbarzadeh, Hua Chai


Hao Zheng, Xinyu Wang, Zehua Qi, Shixuan Sun, Hua Chai, Masoud Akbarzadeh






This research was supported by the PennPraxis and National Science Foundation (NSF) CAREER AWARD (NSF CAREER-1944691- CMMI) to Dr. Masoud Akbarzadeh.