Integrating system dynamics and machine learning for environmental impact analysis of building materials in the...

The construction sector's substantial contribution to climate change has raised significant environmental concerns, particularly regarding building demolitions-a major source of environmental degradation. As a critical phase within the construction lifecycle, the demolition process involves numerous interrelated variables, such as housing units, and demolition tools employed. Therefore, a comprehensive analysis of these factors is crucial for mitigating environmental impact. This study integrates System Dynamics (SD) modeling with the Random Forest (RF) algorithm to analyze the relationships among key variables and assess the overall environmental impacts of different demolition tool combinations (OEIC1, OEIC2, OEIC3, and OEIC4). The RF algorithm validates the SD model, while sensitivity analysis enhances robustness. The RF model shows strong predictive accuracy, with R-squared values of 0.9776 to 0.9990 and mean squared errors between 0.0003 and 0.0356. Among the combinations, OEIC4-comprising a demolition robot and hydraulic splitter-proves to be the most environmentally sustainable, producing impacts up to four times lower than the others. The analysis also finds that shorter demolition durations substantially reduce environmental impact. This integrated framework provides a novel method for assessing demolition-related environmental impacts. This approach contributes to circular economy principles, reduces pollution and waste, and aligns with global climate change mitigation goals.