Abstract

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.

Download paper here

Download preprint here

Finacial support: National Key R&D Program (No. 2023YFC3805800); National Natural Science Foundation of China (No. 52378306)

Financial Sources: