基于BP神经网络的钢-混组合结构PBL剪力键承载力

Shear Resistance of Perfobond Shear Connectors in Steel-concrete Composite Structure Based on BP Neural Network

为实现钢-混组合结构PBL剪力键极限承载力的准确预测,通过分析既有推出试验结果,对PBL剪力键的作用机理及破坏模式进行总结,确定PBL剪力键纵向抗剪承载力的主要影响因素为开孔直径、钢板厚度、混凝土抗压强度、贯穿钢筋直径、钢板屈服强度等。以神经网络理论为基础,选用误差反向传播(BP)神经网络算法模型,选取钢板厚度、开孔直径、贯穿钢筋直径和混凝土抗压强度为输入因子,用C语言编写出基于BP神经网络的钢-混凝土组合结构PBL剪力键极限承载力预测系统。利用国内外既有试验资料对PBL剪力键极限承载力神经网络预测系统进行网络训练及预测验证。结果表明:利用BP神经网络可以建立主要影响因素与PBL剪力键极限承载力之间的非线性映射关系,实现PBL剪力键极限承载力预测。

In order to accurately evaluate ultimate shear resistance of peffobond （PBL） shear connectors in steel-concrete composite structure, the mechanism and breakage mode of PBL shear connectors were investigated based on the existing push-out test results. According to the push-out test results, longitudinal shear resistance of perfobond shear connectors depends on diameter of holes, thickness of steel plate, compressive strength of concrete, diameter of transverse rebar, yield strength of steel plate, etc. Based on neural network theory, choosing thickness of steel plate, diameter of holes, diameter of transverse rebar and compressive strength of concrete as elements in input layer, the back propagation neural networks （BPNN） model was developed for forecasting the ultimate shear resistance of the PBL shear connectors by using C Language. By using domestic and foreign experimental test data, the BPNN forecast system for ultimate shear resistance of PBL shear connectors were trained and verified. It is demonstrated that the nonlinear mapping relationship between ultimate shear resistance and design parameter of PBL shear connectors can be built up by BPNN to accurately forecast the shear resistance.