基于神经网络的钢纤维橡胶混凝土蠕变研究

Research on the Mechanical Properties of Steel Fiber Rubber Concrete Based on Artificial Neural Network

为探究不同橡胶替代率和钢纤维掺量对混凝土长期受力性能的影响,通过弯曲蠕变实验对混凝土小梁施加30%抗折强度的荷载并持续3天,结果表明,钢纤维的加入可以显著改善橡胶混凝土的蠕变劲度。同时基于蠕变实验数据,建立BP神经网络模型对钢纤维橡胶混凝土的蠕变劲度进行预测,并通过遗传蚁群算法(GA-ACO-BP)对其进行优化,平均绝对百分比误差(MAPE)从16.28%降低至3.3%,有效提高了模型的准确性和稳定性。

In order to investigate the effects of different rubber replacement rates and steel fiber admixture on the long term stress performance of concrete,a load of 30%flexural strength was applied to concrete beams by bending creep experiments for 3 days,and the results showed that the addition of steel fiber could significantly improve the creep stiffness of rubber concrete.Meanwhile,based on the creep experimental data,a BP neural network model was established to predict the creep stiffness of steel fiber rubber concrete,and it was optimized by genetic ant colony algorithm(GA-ACO-BP),and the Mean absolute percentage error was reduced from 16.28%to 3.3%,which effectively improved the accuracy and stability of the model.