Machine Learning for Whole-Building Life Cycle Assessment: A Systematic Literature Review

Life Cycle Assessment (LCA) is a methodology to systematically investigating impacts from interactions between environment and human activities. However, the number of parameters and uncertainty factors that characterize built impacts over their full-lifecycle, preclude a broader LCA adoption. This enable faster progress towards reducing building impacts by combining established environmental impact assessment methods with artificial intelligence approaches, such as machine learning (ML) and neural networks. This article will present previous research on ML for LCA of buildings. To achieve this goal, we perform a Systematic Literature Review (SLR). SLR was governed by the question "What are scientific research developed for Architecture, Engineering and Construction (AEC) industry in LCA and ML context?". This SLR was performed in three databases: Scopus, Engineering Village and Web of Science, using keywords: Life Cycle Assessment, Machine Learning, Learning, Building and Neural Network. From SLR, we identified best practices, acquired and developed by other studies, clarifying how to interpret large data sets monitored through advanced analysis to improve LCA. The results showed: (i) number of articles increase in recent years; (ii) the most searched environmental indicators are energy consumption and Global Warming Potential (GWP); (iii) machine learning is mainly used for prediction impacts and; (iv) the most used ML method is Artificial Neural Networks. Advances in LCA and ML field can contribute to calculation and analysis of buildings environmental indicators, as well as can develop and improve LCA methods. The combination of reliable data and ML will produce an unprecedented change in speed and accuracy of LCA.