Prediction of ultimate bearing capacity of RC beams after fire: data-driven method

The ultimate bearing capacity of reinforced concrete (RC) beams after exposure to fire was investigated using numerical simulations, regression fitting and machine learning techniques to examine the thermal and mechanical properties of the beams after high-temperature exposure. A series of fire and static load tests was conducted on seven RC T-beams. Based on the experimental observations and considering the effects of high-temperature concrete spalling and steel–concrete bond degradation, a numerical model was developed to simulate the temperature distribution and structural behaviour of RC T-beams. The accuracy of the numerical model was validated by comparing the cross-sectional temperature profiles and ultimate bearing capacities after fire exposure with experimental results. A dataset of 500 samples was established, with variables including fire exposure time, depth of concrete spalling, spalling area ratio and loading conditions. Regression fitting and machine learning techniques were used to establish predictive formulas and models for estimating the ultimate bearing capacity of RC T-beams after fire exposure. The accuracy of both methods was found to be within 10%.