Thermal and carbonation resistance of tunnel concrete: Experimental evaluation and hybrid ANN–GPR modeling under fire–CO₂ exposure

This study investigates the combined effects of high-temperature fire exposure and CO₂-induced carbonation on the mechanical and durability performance of concrete used in urban tunnel infrastructure. Experimental results indicated up to a 43.5 % reduction in compressive strength after exposure to 1000 °C for 180 min, primarily due to thermal degradation and porosity increase. Carbonation depth increased by 24.8 % under the same conditions, attributed to microcracking and enhanced CO₂ ingress. Microstructural characterization using XRD revealed decomposition of Ca(OH)₂ and partial formation of CaCO₃, confirming mineralogical deterioration. Ultrasonic pulse velocity (UPV) measurements showed strong correlation with both strength loss and carbonation depth, supporting UPV as a reliable non-destructive evaluation method. A hybrid machine learning model combining artificial neural networks (ANN) and Gaussian process regression (GPR) was developed to predict residual compressive strength and carbonation depth based on UPV and exposure parameters. The model achieved high predictive accuracy (R² > 0.93), offering a robust, data-driven framework for rapid structural assessment. These findings provide valuable insights into post-fire durability evaluation and support effective maintenance strategies in tunnel engineering.