Analysis of Concrete Air Voids: Comparing OpenAI-Generated Python Code with MATLAB Scripts and Enhancing 2D Image Processing Using 3D CT Scan Data

The air void system in concrete significantly affects its mechanical, thermal, and frost durability properties. This study explored the use of ChatGPT, an AI tool, to generate Python code for analyzing air void parameters in hardened concrete, such as total air void content (A), specific surface (α), and air void spacing factor (L). Initially, Python scripts were created by requesting ChatGPT-3.5 to convert MATLAB scripts developed by Fonseca and Scherer in 2015. The results from Python closely matched those from MATLAB when applied to polished sections of seven different concrete mixes, demonstrating ChatGPT’s effectiveness in code conversion. However, generating accurate code without referencing the original MATLAB scripts required detailed prompts, highlighting the need for a strong understanding of the test method. Finally, a Python script was applied to modify void reconstruction in 2D images into 3D by stereology, and comparing this with (3D) CT scanner results, showing comparable results.