Embodied carbon from building materials and construction accounts for 11% of the total global carbon emissions. The building industry urgently needs to transition to carbon sequestering structural materials to meet UN climate targets. However, the time-intensive nature of structural design creates a significant bottleneck; In early design stages, sharp deadlines mean engineers will likely stick to tried-and-tested, but environmentally harmful, options like steel and concrete. As a result, to speed-up adoption of more sustainable structural materials, like mass timber, early-stage structural design must be re-imagined. In this whitepaper, we present a novel data-driven structural design process which leverages synthetic data generation, and machine learning predictions to meet this need. We show how this new process informs the development of a user-friendly design tool which allows project stakeholders to quickly generate and compare structural design options and scenarios with respect to cost, embodied carbon and sequestered carbon. To extend capabilities of the tool, we currently work on generating more synthetic data with the use of a supercomputer and integrating Environmental Product Declarations (EPDs) to embodied carbon calculations.

‍
Keywords: Structural design, Embodied Carbon, Machine Learning, Surrogate modeling

Authors: Aleksei Kondratenko, Matthew Tam, Clemens Preisinger, Rutvik Deshpande, Cesar Cheng, Naomi Bachtiar, Miriam Corcuera, Tejas Chavan Camiel Weijenberg, Sayjel Vijay Patel, Maciej Nisztuk

‍

DBF Synthetic Structural Generation for Early-Stage Carbon Evaluation
DBF, Pte ltd Singapore, 2023