Residual thermo-mechanical response of multi-fibre-reinforced ternary blended self compacting concrete: experimental characterization and PINN-based prediction

Self-compacting concrete (SCC) is a highly flowable concrete capable of filling densely reinforced sections under its own weight without vibration or segregation. This study investigates the residual compressive strength, split tensile strength, and static modulus of elasticity of ternary blended SCC containing fly ash (FA) and rice husk ash (RHA), reinforced with steel–basalt and steel–sisal hybrid fibre systems, after exposure to elevated temperatures ranging from 200 to 800 °C. Among the investigated mixtures, the steel–basalt hybrid exhibited the best post-fire mechanical performance, retaining up to 39% higher compressive strength and more than 100% greater modulus of elasticity than the control mix at 800 °C. A Physics-Informed Neural Network (PINN) incorporating soft physics-based constraints derived from concrete thermal damage mechanisms was developed to predict the residual mechanical properties and compared with linear and ridge regression models using five-fold cross-validation. The PINN achieved superior prediction accuracy, with R2 values of 0.94 ± 0.02 for compressive strength, 0.87 ± 0.12 for the static modulus of elasticity, and 0.82 ± 0.12 for split tensile strength. The results demonstrate that integrating physics-based constraints into deep learning improves the prediction of post-fire mechanical properties of hybrid fibre-reinforced SCC, particularly for compression-dominated behaviour.