Evaluation of Durability Performance for Chloride Ingress Considering Long-Term Aged GGBFS and FA Concrete and Analysis of the Relationship between Concrete Mixture Characteristic and Passed Charge Using Machine Learning Algorithm

Abstract

In this study, accelerated chloride diffusion tests are performed on ordinary Portland cement (OPC), ground granulated blast furnace slag (GGBFS), and fly ash (FA) concretes aged 4–6 years. Passed charge is evaluated according to ASTM-C-1202 for 12 mixtures, considering water–binder (W/B) ratios (0.37, 0.42, and 0.47), GGBFS replacement rates (0%, 30%, 50%), and FA replacement rates (0% and 30%). The effects of aged days on passed charge reduction behavior are quantified through repetitive regression analysis. Among existing machine learning (ML) models, linear, lasso, and ridge models are used to analyze the correlation of aged days and mix properties with passed charge. Passed charge analysis considering long-term age shows a significant variability decrease of passed charge by W/B ratio with increasing age and added admixtures (GGBFS and FA). Furthermore, the higher the water–binder ratio in GGBFS and FA concretes, the greater the decrease in passed charge due to aged days. The ML model-based regression analysis shows high correlation when compressive strength and independent variables are considered together. Future work includes a correlational analysis between mixture properties and chloride ingress durability performance using deep learning models based on the time series properties of evaluation data.