Data-driven Embodied Carbon Evaluation of Early Building Design Iterations

In the early design phases, Life Cycle Assessment can assist project stakeholders in making informed decisions on choosing structural systems and materials with an awareness of environmental sustainability through their embodied carbon content; yet embodied carbon is difficult to quantify without detailed design information in the early design stages.In response, this paper proposes a novel data-driven tool, prior to the definition of floor plan layouts to perform embodied carbon evaluation of existing building designs based on a Bayesian Neural Network (BNN) regression.The BNN is built from data drawn from existing floor plans of residential buildings, and predicts material volume and embodied carbon from generic design parameters typical in the early design stage.Users will be able to interact with the tool in Grasshopper or as an online resource, input generic design parameters, and obtain comparative visualizations based on the choice of a construction system and its environmental sustainability in a 'shoebox' interface -a simplified three-dimensional representation of a building's primary spatial units generated with the tool.