Design and durability analysis method of underground concrete materials based on Chat Generative Pre-trained Transformer

Abstract

Chat Generative Pre-trained Transformer (ChatGPT) has been applied in many fields due to its powerful functions, but its application in the field of concrete materials is very limited. Therefore, this paper innovatively applies it to the design and durability analysis of concrete materials. Through 58 questions from shallow to deep, the recognition of ChatGPT-4 Turbo on the durability of underground concrete was explored from the aspects of originality, readability and accuracy, and the feasibility of applying it to the analysis and identification of underground concrete cracks, mix proportion design, material composition optimization and other fields was verified by experiments. The results show that ChatGPT-4 Turbo can accurately understand the basic knowledge of underground concrete materials. Based on its multimodal ability and updated iterative algorithm, it also yields good results in concrete image recognition, mix design, material composition optimization, and so on. It can help scholars quickly and accurately understand the relevant knowledge of underground concrete and provide a reference for the design, optimization, and performance prediction of concrete materials in practical engineering. Finally, this paper also introduces some limitations of ChatGPT-4 Turbo and possible future research directions and approaches.