基于机器学习的高温损伤后UHPC残余抗压强度预测  

Machine learning based prediction of residual compressive strength of UHPC after high temperature damage

作　　者：欧阳利军[1] 容文照 谢冰清 丁斌[2] OUYANG Lijun;RONG Wenzhao;XIE Bingqing;DING Bin(School of Environment and Architecture,University of Shanghai for Science and Technology,Shanghai 200093,China;School of Civil Engineering&Architecture,Wenzhou Polytechnic,Wenzhou 325035,China)

机构地区：[1]上海理工大学环境与建筑学院,上海200093 [2]温州职业技术学院建筑工程学院,浙江温州325035

出　　处：《混凝土与水泥制品》2025年第2期1-7,共7页China Concrete and Cement Products

基　　金：国家自然科学基金项目(51708349);浙江省自然科学基金项目(LY20E080017);温州市科协服务科技创新项目(kjfw53)。

摘　　要：以水胶比、m_(硅灰)∶m_(水泥)、m_(粉煤灰)∶m_(水泥)、m_(石英砂)∶m_(水泥)、m_(石英粉)∶m_(水泥)、钢纤维掺量、PP纤维掺量、常温标准养护、热水养护、干热养护和加热温度11个影响因素为输入变量,建立了BP神经网络、麻雀搜索算法优化人工神经网络、遗传算法优化人工神经网络(GA-BP)和支持向量机回归四种模型,并对高温损伤后的超高性能混凝土(UHPC)残余抗压强度进行了预测。结果表明:与基于试验经验的计算模型预测结果相比,以上4个机器学习模型的预测精度较高,误差基本控制在15%以内,其中,GA-BP模型的预测结果最优,R^(2)达到0.949。By using 11 influencing factors of water-cement ratio,m_(silica):mcement,m_(fly ash):m_(cement),m_(sand):m_(cement),m_(quartz powder):m_(cement),steel fiber content,PP fiber content,room temperature standard curing,hot water curing,dry heat curing and heating temperature as input variables,four models of a BP neural network,sparrow Search Algorithm Optimized Artificial Neural Network,Genetic Algorithm Optimized Artificial Neural Network(GA-BP),and Support Vector Machine Regression were established to predict the residual compressive strength of ultra-high performance concrete(UHPC)after high temperature damage.The results show that compared with the prediction results of experimental experience calculation models,the four machine learning models have higher prediction accuracy,with an error basically within 15%.Among them,the GA-BP model has the best prediction result,with the R^(2) of 0.949.

关 键 词：超高性能混凝土 残余抗压强度 高温损伤 机器学习 预测 

分 类 号：TU528.1[建筑科学—建筑技术科学]