{"id":9396,"date":"2025-10-16T10:16:13","date_gmt":"2025-10-16T08:16:13","guid":{"rendered":"https:\/\/ingenius.ecoledesponts.fr\/?p=9396"},"modified":"2025-10-27T10:30:31","modified_gmt":"2025-10-27T09:30:31","slug":"artificial-intelligence-serving-eco-design","status":"publish","type":"post","link":"https:\/\/ingenius.ecoledesponts.fr\/en\/articles\/artificial-intelligence-serving-eco-design\/","title":{"rendered":"Artificial intelligence serving eco-design"},"content":{"rendered":"\n\n\n
\"\"
Building structures generated by artificial intelligence, with a Climate Change Index of 200 kg of CO\u2082-eq\/m\u00b2 and a total floor area of 2,000 m\u00b2. (Credit : Maxime Pollet)<\/figcaption><\/figure>\n\n\n\n

<\/p>\n\n\n\n

Life Cycle Assessment (LCA) makes it possible to compare the potential environmental impacts of different construction choices, but its application during the design phase remains limited in practice. Designers have to quickly explore a wide range of construction scenarios and do not necessarily have the time and resources required to model a building\u2019s life cycle. To overcome this challenge, a research project has been conducted at the Navier Laboratory, under the supervision of Olivier Baverel and Ad\u00e9la\u00efde Feraille. Funded by the CNRS since May 2025, this project explores the potential uses of artificial intelligence methods. This approach could make it possible to predict a building\u2019s environmental impacts in just a few milliseconds, and even to generate multiple building proposals that align with criteria defined by the designer. If proven effective, these methods could pave the way for the integration of LCA into the design process.<\/p>\n\n\n\n

Design and Life Cycle Assessment<\/h2>\n\n\n\n

The design phase is a key step in the building construction process. Architects and engineers put forward proposals, draw up plans, and carry out technical studies. The goal is to transform the client\u2019s needs into an actual project that is technically feasible, and compliant with current regulations. The decisions made during the design phase are therefore crucial in determining the building\u2019s overall environmental impact. However, many of these decisions cannot be based on a rigorous assessment of environmental performance, due to a lack of available information.<\/p>\n\n\n\n

LCA is a method for estimating the potential impacts associated with each stage of a building\u2019s life cycle, from the extraction of raw materials to final deconstruction. These environmental impacts are categorized using indicators that quantify various effects, such as greenhouse gas emissions, water consumption, and carcinogenic toxicity. LCA is a powerful tool for comparing alternative construction choices because it can distinguish both the stages of the life cycle and the types of impacts. Recently, the RE2020 regulation<\/a> has made LCA a standard requirement in the design phase of buildings. However, the processes involved in collecting and organizing LCA data are still time-consuming, which limits the exploration of multiple construction solutions and hinders its use as a decision-support tool. As a result, in practice, and despite the introduction of RE2020, LCA is often conducted late in the design process, once a specific building design has already been selected.<\/p>\n\n\n\n

Artificial intelligence and eco-design<\/h2>\n\n\n\n

            Launched in May 2025, this research project aims to develop artificial intelligence-based solutions that allow designers to rapidly explore a wide range of building typologies, while simultaneously considering their environmental impacts.<\/p>\n\n\n\n

Artificial intelligence is a broad field, but most recent advances stem from one of its sub-fields: machine learning. This involves modeling causal relationships between different variables, not by applying principles known to the user, but rather by inferring patterns from datasets. For example, if we want to model the behavior of a beam under load, a machine learning approach would not use a traditional engineering approach. Instead, it would use a database containing many examples of beam performance, and use statistical inference to predict how other beams would behave. Such an approach can be useful in two types of scenarios:<\/p>\n\n\n\n