Machine Learning Strategies for early prediction of structural embodied emissions

The building sector’s pursuit of carbon neutrality has led to a surge in regulatory requirements
within structural engineering. To comply with environmental policies and mitigate the significant impact
of structural design on Greenhouse Gas (GHG) emissions, structural engineers must adapt their method-
ologies. Thus, there is a pressing need to familiarize them with Life Cycle Analysis (LCA) principles and
integrate efficient LCA tools into their workflow, reducing reliance on external environmental practices.
This research, conducted in collaboration with Bollinger + Grohmann (B+G), addresses the challenges
faced by structural engineers and explores key changes to reduce project environmental impacts without
increasing workload. A primary challenge is the diverse array of working habits and tools within the same
office, requiring tailored solutions. Additionally, projects’ dynamic nature and regular changes, often de-
fers LCAs to final stages, limiting design changes. The solutions implemented as part of the overarching
LCA strategy for B+G encompass the development of user-friendly LCA plug-ins seamlessly integrated
into in-house tools, alongside the automation of in-house LCA data archiving. In particular, this research
has catalyzed the development of an early design tool providing real-time estimation of embodied GHG
emissions, enabling comparative analyses of buildings, explanation of feature impacts, and verification
of computed results. As an alternative to conventional LCA calculation, the approach (Fig 1) followed to
create this tool leverages knowledge from existing building’s databases, and uses contextual and descrip-
tive data, such as structure typology and location, to approximate the total embodied GHG emissions.
Based on machine learning techniques, the methodology is rendered adaptable to diverse database con-
tents, as evidenced by its evaluation across three distinct databases and its proof of concept validation
using in-house B+G data. Based on challenges and opportunities encountered in daily professional prac-
tice, this study aims to assist structural engineering practices in design and decision-making, thereby
bridging data and knowledge gaps in the building industry. Its broader applications could inform envi-
ronmental regulations, policies, and mitigation strategies, facilitating a transition towards low-emission,
resilient, and sustainable built environments.