What is it about?
Physics-informed neural networks have become relevant lately in many areas of civil engineering. Would they have potential too for structural design of reinforced concrete structures where constructability issues are as relevant as the physics behind? In this paper we present a PIGNN (CPyRO-GraphNet-Beams) for surrogate-assisted optimization design of rebar in concrete beams. We demonstrated that by considering the mechanics of materials through the governing PDE of beams, in addition to labeled data, better estimations of optimum rebar designs can be obtained, than with plain GNNs. - We implemented and tested the whole framework with a multi-objective (rebar weight vs constructability) optimization design algorithm. Here, the assistance of our model generated better optimum Pareto fronts of rebar solutions with less material and larger constructability. - To efficiently measure the constructability we propose a holistic constructability score model that considers the whole supply chain of rebar design, from rebar cutting and bending all the way to rebar assembly and placing.
Featured Image
Photo by Yevgeniy Mironov on Unsplash
Why is it important?
To design optimally the rebar of RC structures is computationally expensive. AI is necessary to accelerate the process, specially for large RC buildings and structures. Large amounts of data are required to efficiently train these models and obtain an acceptable accuracy of estimations of optimum rebar quantities. At the end, any amount of error of these models is translated in heavier rebar designs, sacrificing sustainability. - By aiding AI models with the phisics behind rebar design, higher accuracy can be obtained, translated at the end in more sustainable outputs. - But what about the constructability and practicality of these final rebar design outputs? To account for this important issue we implemented an additional GNN for constructability-awareness, as a pretrained entity to feed the main PIGNN.
Read the Original
This page is a summary of: Constructability-aware Physics-Informed Graph Neural Networks for surrogate-assisted optimization design of rebar in concrete beams, Automation in Construction, February 2026, Elsevier,
DOI: 10.1016/j.autcon.2026.106760.
You can read the full text:
Resources
Contributors
The following have contributed to this page
