AI-Driven Prediction Of Mechanical And Durability Performance Of Autoclaved Aerated Concrete Under Prolonged Water And Moisture Exposure

Abstract

Autoclaved aerated concrete (AAC) is used in many light wall and partition structures because of its low density and good thermal insulation. However, because of its porous nature water can enter the material and can affect the strength and long-term behavior of the material. AAC specimens will be made in a controlled mixture and exposed to different moisture conditions in different periods. Density, water absorption, moisture content, compressive strength, splitting tensile strength, ultrasonic pulse velocity and dimensional change will be measured at different ages.Selected specimens will also be examined using scanning electron microscopy, X-ray diffraction and Fourier-transform infrared spectroscopy to understand changes in the internal structure. The experimental results will be used to develop a database for machine-learning prediction. Multiple linear regression, random forest, support vector regression, artificial neural network and XGBoost will be compared. Model performance will be assessed using R2, RMSE and MAE. SHAP analysis will then be used to determine which material and exposure variables have the greatest effect on the predicted properties. Finally, selected predictions will be checked through additional laboratory tests. The study is intended to provide a practical method for estimating the performance of AAC subjected to prolonged moisture exposure.