Abstract:

Aiming at the bond?slip behavior of reinforced concrete, the finite element model of reinforced concrete bond?slip based on cohesion model was constructed by ABAQUS finite element software. The mesh sensitivity and cohesion parameter sensitivity of the simulation model were explored by energy and load?displacement curves. Aiming at the problem of bond strength of reinforced concrete, a nonlinear autoregressive exogenous network (NARX) was developed to predict the load?displacement curve for reinforced concrete by creating 20 sets of data with the variables of bond length, reinforcement diameter, and loading method. The study shows that the mesh size of 6 mm provides an ideal balance between prediction accuracy and computational cost. Based on the sensitivity of the finite element prediction results, the cohesive parameters are in the sequence of damage initiation strength, fracture energy, and stiffness. The NARX with the prediction accuracy of 99.6% is promising to replace time?consuming numerical simulations and experimental works to achieve an efficient and accurate prediction of the bond strength of reinforced concrete. Such an efficient and accurate prediction method provides a novel and convenient methodology of predicting and designing the bond strength of reinforced concrete.

Key words: Reinforced concrete bond?slip behavior, Cohesive parameters, Mesh size, Sensitivity analysis, NARX

摘要：

针对钢筋混凝土黏结?滑移行为，利用ABAQUS有限元软件，构建了基于内聚力模型的钢筋混凝土黏结?滑移有限元模型，通过能量和荷载?位移曲线探究了仿真模型网格敏感性以及内聚力参数敏感性。针对钢筋混凝土黏结强度问题，建立基于非线性自回归动态神经网络模型（NARX）的预测模型，以黏结长度、钢筋直径和加载方式为变量，建立20组数据对钢筋的荷载?位移曲线进行了预测。结果表明，当网格尺寸为6 mm时，可以较理想地平衡预测精度与计算成本；有限元预测结果对内聚力参数的敏感性由强到弱依次为损伤起始强度、断裂能和刚度；所建立的NARX预测精度达到99.6%，有潜力代替量大且耗时的数值模拟和物理试验，实现对钢筋混凝土黏结强度的高效准确预测，为钢筋混凝土黏结强度的预测和设计提供新的便捷途径。

关键词: 钢筋混凝土黏结?滑移行为, 内聚力参数, 网格尺寸, 敏感性分析, NARX