Building energy conservation and emission reduction are crucial in addressing global climate change. High-performance insulated building envelopes can significantly reduce energy consumption over a building’s lifecycle. However, few studies have systematically analyzed carbon reduction potential through a life cycle assessment (LCA), incorporating case studies and regional differences. To address this, this study establishes an LCA carbon emission calculation model using Building Information Modeling (BIM) technology and the carbon emission coefficient method. We examined four residential buildings in China’s cold regions and hot summer–cold winter regions, utilizing prefabricated concrete sandwich insulation exterior walls (PCSB) and autoclaved aerated concrete block self-insulating exterior walls (AACB). Results indicate that emissions during the operational phase account for 75% of total lifecycle emissions, with heating, ventilation, and air conditioning systems contributing over 50%. Compared to AACB, PCSB reduces lifecycle carbon emissions by 18.54% and by 20.02% in hot summer–cold winter regions. The findings demonstrate that PCSB offers significant energy-saving and emission-reduction benefits during the construction and operation phases. However, it exhibits higher energy consumption during the materialization and demolition phases. This study provides a practical LCA carbon calculation framework that offers insights into reducing lifecycle carbon emissions, thereby guiding sustainable building design.
With the rapid advancement of artificial intelligence, large language models (LLMs) such as ChatGPT have garnered increasing attention in the field of text generation. However, when applied to patent drafting in the architectural domain, these models often produce verbose texts and omit critical elements, leading to issues such as limited claim scope and insufficient disclosure in the specifications. To address these challenges, this study conducted experimental research on the application of AI technologies, including GPT-4o, in the drafting and publication of patents related to prefabricated integral buildings. The results indicate that GPT-4o performs well in generating claims and technical descriptions. Nevertheless, problems such as repetitive descriptions of technical achievements, ambiguity in result interpretation, and a lack of detailed implementation methods were observed during the generation process. Therefore, we recommend incorporating expert review mechanisms when using GPT-4o for patent drafting, to enhance the accuracy and professionalism of the output. Moreover, enriching training datasets and enforcing a rigorous textual review workflow can further optimize model outputs, aligning them more closely with the high standards required in patent documentation.