Durability evaluation of GFRP rebars in harsh alkaline environment using optimized tree-based random forest model

Abstract

GFRP bars reinforced in submerged or moist seawater and ocean concrete is subjected to highly alkaline conditions. While investigating the durability of GFRP bars in alkaline environment, the effect of surrounding temperature and conditioning duration on tensile strength retention (TSR) of GFRP bars is well investigated with laboratory aging of GFRP bars. However, the role of variable bar size and volume fraction of fiber have been poorly investigated. Additionally, various structural codes recommend the use of an additional environmental reduction factor to accurately reflect the long-term performance of GFRP bars in harsh environments. This study presents the development of Random Forest (RF) regression model to predict the TSR of laboratory conditioned bars in alkaline environment based on a reliable database comprising 772 tested specimens. RF model was optimized, trained, and validated using variety of statistical checks available in the literature. The developed RF model was used for the sensitivity and parametric analysis. Moreover, the formulated RF model was used for studying the long-term performance of GFRP rebars in the alkaline concrete environment. The sensitivity analysis exhibited that temperature and pH are among the most influential attributes in TSR, followed by volume fraction of fibers, duration of conditioning, and diameter of the bars, respectively. The bars with larger diameter and high-volume fraction of fibers are less susceptible to degradation in contrast to the small diameter bars and relatively low fiber's volume fraction. Also, the long-term performance revealed that the existing recommendations by various codes regarding environmental reduction factors are conservative and therefore needs revision accordingly.