基于PCA-BP原理的混凝土碳化深度预测  

Prediction of Carbonation Depth of Concrete Based on PCA-BP Principle

作　　者：甘海龙[1] 郭容宽[1] Gan Hailong;Guo Rongkuan(Guangxi Technological College of Machinery and Electricity,Nanning Guangxi 530007,China)

机构地区：[1]广西机电职业技术学院,南宁530007

出　　处：《科技通报》2019年第12期144-149,154,共7页Bulletin of Science and Technology

基　　金：2017年度广西高校中青年教师基础能力提升项目(项目编号:2017KY075)的研究成果。

摘　　要：混凝土碳化深度是钢筋混凝土结构耐久性评估的重要参数,影响混凝土碳化深度的因素主要有水灰比、水泥用量、混凝土抗压强度、碳化时间、水泥强度、温度与湿度。基于以上7个参数,并结合BP神经网络较好的预测性,以及主成分分析(PCA)能消除自变量间的多重共线性和降低输入数据维度的特点,建立了基于PCA-BP神经网络的混凝土碳化深度预测模型。以30组实测数据为例,对7个影响因素进行主成分分析,最终降为4个主成分,进而将其作为BP神经网络的输入因子,对混凝土碳化深度进行了预测。结果表明:PCA-BP神经网络预测误差低,实现了对混凝土碳化深度的较准确预测,PCA-BP神经网络模型为混凝土碳化深度预测提供了一种科学、可靠的方法。The carbonation depth of concrete is an important parameter to evaluate the durability of reinforced concrete structure.The main factors affecting the carbonation depth of reinforced concrete are water-cement ratio,cement content,compressive strength of concrete,carbonation time,cement strength,temperature and humidity.Based on the above seven parameters and combining with BP neural network,the characteristics of multiple collinearity among independent variables and the dimension of input data can be eliminated by principal component analysis(PCAs).The prediction model of concrete carbonation depth based on PCA-BP neural network is established.Taking 30 groups of measured data as an example,7 influencing factors were analyzed by principal component analysis and finally reduced to 4 main factors.The carbonation depth of concrete was predicted by using the component as the input factor of BP neural network.The results show that the prediction error of the PCA-BP neural network is low,and the PCA-BP neural network model provides a scientific and reliable method for predicting the carbonation depth of concrete.

关 键 词：碳化深度 主成分分析法 BP神经网络 预测模型 

分 类 号：TU528.517[建筑科学—建筑技术科学]