摘要

生成式人工智能正驱动工程建设行业从数字化向智能化跨越。深度剖析生成式AI大模型在工程建设行业全生命周期的应用现状与趋势，创新性地提出基于信息熵流变的输入/输出(I/O)任务分类框架。即内容生成类(I/O扩张)可实现从抽象意图到具象建筑方案、三维建筑信息模型及施工文档的高效生成；解析推理类(I/O均衡)能打破数据孤岛，实现图纸与BIM模型等工程资产的语义级解析、跨模态比对与逆向重构；评估决策类(I/O压缩)则在图纸智能合规审查、施工进度风控及运维能源调度中展现出显著决策优势。研究显示，生成式AI大模型的工程建设应用正全面迈入高效场景落地阶段，将持续助力行业的效率与效益提升。

Generative artificial intelligence is driving the engineeringe construetion industry to leap from digitization to intelligence. The application situation and trend of generative AI large model in the full life cycle of engineering construction is deeply analyzed, and an input/output (I/O) task classification framework based on flow of information entropy is innovatively proposed. That is, the content generation class (I/O expansion) can realize the effieient generation from abstract intention to conerete arehitectural scheme, three dimensional arehiteetural information model and construction document. The analytical reasoning class(I/O equilibrium) can break the data island and realize the semantic level analysis, cross-modal comparison and reverse reconstruction of engineering assets such as drawings and BIM models. The evaluation decision class(I/O compression) shows significant decision-making advantages in drawing intelligent compliance review, construetion schedule risk control and operation and maintenance energy scheduling. The research shows that the engineering construction application of the generative AI large model is entering the stage of efficient scene landing in an all-round way, which will continue to improve the efficiency and benefit of the industry.

论文下载链接

预印本下载链接

基金资助: 国家重点研发计划资助项目（2023YFC3805800）； 国家自然科学基金资助项目（52378306）

相关项目资助情况: