基于神经网络的HPFL加固层与混凝土粘结强度预测.pdf

基于神经网络的 HPFL 加固层与混凝土粘 结强度预测# 黄华，施明君，刘鸣* 5 （长安大学建筑工程学院，西安 710061） 摘要：在 243 个测点的正拉粘结强度试验和 24 个测点的剪切粘结强度试验的基础上，提取 抹灰龄期、界面粗糙度、混凝土和砂浆强度、修补方位等特征参数，建立了预测 HPFL 加固 层与混凝土粘结强度的 BP 人工神经网络模型。采用训练好的 BP 神经网络对 HPFL 加固层 与混凝土之间的粘结强度进行了预测，并与实测值进行了对比。其中，正拉粘结强度预测值 10 与试验值之比的平均值为 1.056，标准差为 0.057 ；剪切粘结强度预测值与试验值之比的平均 值平均为 0.988，标准差为 0.127 。结果表明：预测值与试验值符合良好，利用BP 神经网络 对 HPFL 加固层与混凝土之间的粘结强度进行预测是可行的。 关键词：BP 神经网络；HPFL 加固层；正拉粘结强度；剪切粘结强度 中图分类号：U375 15 Prediction of bond strength between HPFL Reinforced Layer and concrete Based on BP Neural Network HUANG Hua, SHI Mingjun, LIU Ming (School of Civil Engineering,Changan University, Xian 710061) 20 Abstract: Based on the experiments about the tensile bond strength of 243 observation points and the shear bond strength of 24 observation points, characteristic parameters of bond strength between HPFL reinforced layer and concrete, i. e. curing age, interface roughness, the strength of concrete and mortar, position of the repaired interface were extracted, and the BP neural network models were developed to predict the bond strength between HPFL reinforced layer and concrete. 25 The bond strength values were predicted using the trained BP neural network and were compared with the measured values. The average ratio between the calculating values and test results of the tensile bond strength is 1.056, and the standard error is 0.057. The average ratio between the calculating values and test results of the shear bond strength is 0.988, and the standard error is 0.127. Results show that the calculating values are more secure, and using BP neural network to 30 predict the bond strengt