Application of artificial intelligence in predicting the residual mechanical properties of fiber reinforced concrete (FRC) after high temperatures

Abstract

The practical application of Artificial Intelligence (AI) approaches in estimating the mechanical properties of fiber-reinforced concrete (FRC) subjected to high temperatures was initiated by developing a dataset including concrete mixtures, geometrical and mechanical properties of fiber, and temperatures. The dataset contains compressive strength, tensile strength, and modulus of elasticity of concrete with two distinct types of fiber, i.e., steel and polypropylene with an extensive range of temperatures from 100 to 1200 ℃. The dataset determined the gaps in the literature and showed FRC with high fiber aspect ratios and higher content of fiber (>0.8%) are not studied enough. The AI-based models showed that steel fiber-reinforced concrete (SFRC) has higher residual compressive strength compared with polypropylene fiber-reinforced concrete (PFRC). An increase in steel fiber diameter and length resulted in a higher compressive strength ratio at all temperatures. Moreover, higher PP fibers content and longer PP fibers decrease the rate of tensile strength degradation. A large probabilistic analysis was performed, and it showed that the failure probability (Pf) for PFRC is independent of W/B ratios and is almost 50%;. In contrast, for SFRC, W/B ratios between 0.4 to 0.5 have lower failure probability in comparison with other W/B ratios. Moreover, Pf for both fibers at a W/B ratio of 0.5 is almost independent of fiber content. Furthermore, the critical temperature resulting in failure (defined as the residual strength ratio less than 0.5) for SFRC and PFRC is 550 ℃ and 430 ℃, respectively.