基于BP-ANN与RBF-ANN的钢筋与混凝土黏结强度预测模型研究

Research on prediction model of bond strength between reinforcement and concrete based on BP-ANN and RBF-ANN

为研究神经网络对钢筋与混凝土黏结强度的预测能力以及神经网络的输出性能,基于大量的试验数据,提出一种基于改进神经网络的变形钢筋与混凝土黏结强度预测模型,对混凝土结构的研究与实际工程应用均有着重要的意义。收集290组黏结锚固试验数据,引入基于反向传播人工神经网络(BP-ANN)与径向基函数神经网络(RBF-ANN)算法,揭示混凝土强度、保护层厚度、钢筋直径、锚固长度及配箍率对变形钢筋与混凝土黏结性能的影响规律,建立基于改进神经网络算法的钢筋与混凝土黏结强度预测模型。对比分析不同数据预处理方法和训练神经元个数对建议模型预测结果的影响,评估各经典模型与建议模型的预测精度和离散性,提出临界锚固长度计算公式。结果表明:BP-ANN预测值与试验值比值的均值、标准差及变异系数分别为1.009、0.188、0.86,其预测精度略高于RBF-ANN;建议模型能够更准确、更稳定地预测钢筋与混凝土的黏结强度,该方法为解决钢筋与混凝土黏结问题提供了新思路。

An investigation into the predictive capabilities of neural networks for assessing the bond strength between steel bars and concrete was conducted and the output performance of neural networks was assessed.Utilizing a vast array of experimental data,this research introduced an advanced neural network-based model for predicting the bond strength in deformed steel bars and concrete,a critical consideration for both academic research and practical engineering applications in concrete structures.The study analyzed data from 290 bond anchorage tests,employing back propagation artificial neural network(BP-ANN)and radial basis function neural network(RBF-ANN)algorithms.These algorithms were used to ascertain the influence pattern of concrete strength,protective layer thickness,reinforcement diameter,anchorage length,and stirrup ratio on the bond performance between deformed steel bars and concrete.Additionally,the study established a prediction model based on an enhanced neural network algorithm,evaluating the impact of various data preprocessing techniques and neuron training counts on the model's predictive results.Comparative analyses between the proposed model and traditional models were performed,focusing on predictive accuracy and variability.The research also introduced a formula for calculating critical anchorage length.Findings indicated that the BP-ANN model,with a mean,standard deviation,and coefficient of variation of 1.009,0.188 and 0.86,respectively for the predicted-to-experimental value ratio,slightly outperformed the RBF-ANN in terms of prediction accuracy.This proposed model demonstrated enhanced precision and stability in predicting the bond strength between steel bars and concrete,offering novel insights into addressing bonding issues between steel bars and concrete.