Steel fiber blending can effectively improve the performance of recycled concrete, but the change rule of compressive strength under different loading conditions is complicated. This paper investigates the compressive performance of steel fiber reinforced recycled concrete under different loading conditions and establishes a prediction model based on BP neural network. The study selected apparent density of recycled coarse aggregate, water absorption, steel fiber parameters, cement dosage, water-cement ratio, sand rate and recycled coarse aggregate substitution rate as input parameters, and slump, 28d compressive strength and elastic modulus as output parameters. The model was trained by 7-20-3 network structure with trainlm function. The results show that: the prediction accuracy of the constructed BP neural network model is high, and the overall root-mean-square error is only 0.8671%, which has a good generalization ability; when the water-cement ratio is 0.66, the overall compressive strength of concrete decreases with the increase of the substitution rate of the recycled coarse aggregate; the compressive strength of granular shaping recycled coarse aggregate concrete is higher than that of the simple crushed recycled coarse aggregate concrete; the simple crushed steel fiber recycled concrete has the optimum water-to-cement ratio (0.45) was higher than that of particle shaped steel fiber recycled concrete (0.4); the compressive strength of particle shaped steel fiber recycled concrete was the greatest within the range of water-tocement ratios from 0.35 to 0.4. The model reliably reflects the influence of steel fiber admixture, recycled aggregate properties and water cement ratio on the compressive strength of concrete, which provides a scientific basis for the actual proportion design.
