Progressive prediction embodied carbon emissions about stages of schematic design with machine learning

The construction industry for 40% of global carbon emissions (WBCSD) contributes to greenhouse gas emissions. It is essential to tackle global climate change and the transition to a low -carbon society. After 2023 Study on carbon emissions from buildings and urban infrastructure in ChinaThe total carbon emissions from housing construction in China reached 4.07 billion tons of CO₂ in 2021, with embodied carbon, which performed 41.8% of the total amount (China, 2023). Embodied carbon includes emissions from material extraction, production, transport and construction that occur in the early stages of the building life cycle (Keyhani et al., 2023). In contrast, the operational carbon comes from energy consumption during the operation of a building, including heating, cooling and lighting (IBN-Mohammed et al., 2013). Recent examinations show that surgical carbon is a significant part of the total emissions of a building of approx. 28%, while embodied carbon is about 11%(Keyhani et al., 2023). However, the importance of the embodied carbon increases, especially when buildings become more energy -efficient. In view of the use of operating energy due to efficiency improvements, the relative influence of embodied carbon is more pronounced, which leads to a greater attention to his management and reduction (Neves Mosquini et al., 2022; MF Victoria et al., 2018).

Several studies have shown that decisions made during the design phase are decisive for the embodied carbon of a building, with 70% of these decisions influence the sustainability of the building (Alwan et al., 2022). For example, the architectural redesign during this phase can lead to a reduction of over 30% in embodied carbon (Häkkinen and Kuittinen et al., 2015), while early decisions can lead to variations of up to 15% in emissions (Hens et al., 2021). The schematic design phase lays the basis in the early construction phases because it defines important design parameters. Some studies emphasize that design changes at the beginning of the process have a significant impact on project performance, while in later stages only minor adjustments are possible (Abdelmaged et al., 2020; Orr et al., 2020; Z. Wang et al., 2021). For example, the structural design of a building in the preliminary design phase reflects that of the schematic design phase (Torabi et al., 2024). Therefore, the schematic design phase is crucial for reducing the embodied carbon of a building.

A precise embodied carbon measurement is crucial to make well -founded decisions during the schematic design phase. Currently, the environmental impacts of buildings are mainly assessed based on the framework for life cycle evaluation (LCA), which includes process-based (P-LCA) (Y. Zhang et al., 2019), input-output (IO-LCA) (Leontief, 2018) and Hybrid LCA (H-LCA) (SUH, 2004). While the process-based analysis is based on detailed inventory data to estimate emissions, the input output method uses the data of the economic sector to estimate the upstream emissions (Dixit, 2017). Hybrid methods that combine the strengths of both approaches provide more precise results, but are more complex and time -consuming (Cang, Yang et al., 2020). It is known that conventional LCA approaches are time-consuming and expensive. Building information modeling (BIM) as a digital representation of a building facilitates the integration of different data sources and modeling techniques. For example, Cang, Luo et al. (2020) proposed a BIM-based calculation method with “building units” in order to evaluate embodied carbon quantitative, alvi et al. (2023) examined BIM-based detection systems that can estimate carbon dioxide emissions and energy consumption. However, BIM-based embodied carbon calculations depend on the level of detail of the model, which is often not sufficient in early construction phases due to indefined materials. Hollberg et al. (2016) found that BIM can introduce a considerable complexity and restrict its use in small projects and early stages of construction design. Therefore, the challenge remains to develop a simplified and reliable method to estimate the carbon during the schematic design phase, which means that architects can make well -founded decisions before the design details are completed.

Most predictive models for carbon emissions for early construction phases are based on parameters that are determined in later design phases, which restricts the ability to adapt the design parameters to reduce carbon emissions. For example, Cang, Yang, et al. (2020) and S. Su et al. (2024) focused on the prediction of emissions based on important building materials that are quantified after the construction level. Other studies (El Hafdaoui et al., 2023; Fenton et al., 2024; M. Victoria et al., 2015; MF Victoria et al., 2018; Zhixing et al., 2014) have determined predictive models using design parameters via the schematic phase such as structural components, completion of quality and service quality. In addition, most studies use different variables to predict embodied carbon emissions during the schematic design phase. For example, X. Zhang, Chen et al. (2024) Identified construction parameters such as building height, structural shape, seismic fastening intensity, type of delivery, geographical region and material costs. Baek et al. (2013) used the design parameters, floor surface per standard, floor height and number of rooms. Similarly, Y. Liu et al. (2024) Used parameters such as building floor, building surface, number of construction rooms, number of primary rooms, window conditions and external processing materials. In previous studies, however, the design parameters were not changed from the architect's perspective, during the schematic design phase. Parameters such as geographical location and material costs are fixed background factors that architects cannot change and limits their ability to optimize these factors for reducing carbon emissions.

The second section of this article describes the methodology, Section 3 contains the research results and discusses extended research. The 4th section shows the conclusions, restrictions and future research directions.