基于BP神经网络的既有建筑混凝土强度预测

Predition of Concrete Strength of Existing Buildings Based on BP Neural Networks

在分析检测数据的基础上,提取了结构服役时间、结构建造时间、结构检测时间、混凝土设计强度和混凝土碳化深度等特征参数,建立了预测既有建筑混凝土强度退化的人工神经网络模型。采用动量法和自适应调整法改进了BP算法;采用训练好的BP神经网络对既有混凝土强度最小值和混凝土强度最大值进行了预测,并与实测值进行了对比。结果表明:利用BP神经网络对既有建筑混凝土强度退化进行预测是可行的,该研究成果可为既有建筑大面积的抗震性能普查提供参考。

Based on the test data analysis method,characteristic parameters of the existing buildings,i.e.service time,construction time,in-situ inspection time of structure,design value of concrete strength,and carbonation depth of concrete were extracted,and the artificial neural network model was developed to predict the degradation of concrete strength of the existing buildings.The back propagation（BP） algorithm was improved by using the momentum method and adaptive adjustment method.Both minimum and maximum values of concrete strength were predicted using the trained BP neural network and were compared with the measured values.Results show that using BP neural network to predict the degradation of concrete strength of existing buildings is feasible.Results of this study can provide references for the existing building seismic performances of large area surveys.