基于人工神经网络的碱矿渣混凝土强度预测  

Alkali-activated slag concrete strength prediction based on artificial neural network

作　　者：王雪芳[1,2] 刘豪杰 马世龙 沈妍燃 吴文达[1,2] 许利惟 张晓平 WANG Xuefang;LIU Haojie;MA Shilong;SHEN Yanran;WU Wenda;XU Liwei;ZHANG Xiaoping(School of Advanced Manufacturing,Fuzhou University,Quanzhou 362251,Fujian,China;College of Civil Engineering,Fuzhou University,Fuzhou 350108,China;College of Civil Engineering,Fujian University of Technology,Fuzhou 350118,China;School of Electricaland Control Engineering,North China University of Technology,Beijing 100144,China)

机构地区：[1]福州大学先进制造学院,福建泉州362251 [2]福州大学土木工程学院,福建福州350108 [3]福建工程学院土木工程学院,福建福州350118 [4]北方工业大学电气与控制工程学院,北京100144

出　　处：《南昌大学学报（工科版）》2023年第4期365-369,408,共6页Journal of Nanchang University(Engineering & Technology)

基　　金：泉州市科技计划项目(2022C012R)。

摘　　要：强度是混凝土的重要性能指标,碱矿渣混凝土的强度不仅与水胶比有关,还与碱激发剂相关,故而传统的硅酸盐水泥混凝土强度预测模型不适用于碱矿渣混凝土。为了能够精准预测碱矿渣混凝土的强度,减少试验次数,节约试验资源,将改进BP算法及主成分分析法(PCA)运用于人工神经网络模型,结果显示:相对于T-BP网络模型,基于碱矿渣混凝土构建的PCA-BP神经网络强度预测模型的训练速度更快、训练误差更小、精准度更高,平均相对误差只有3.81%,能够达到理想的预测效果,为有效预测碱矿渣混凝土强度提供了新的途径。Strength is an important performance index of concrete,the strength of alkali-activated slag concrete is not only related to the water-cement ratio,but also to the alkali exciter,so the traditional strength prediction model for silicate cement concrete is not applicable to alkali-activated slag concrete.In order to accurately predict the strength of alkali slag concrete,reduce the number of tests and save test resources,the improved BP algorithm and principal component analysis were applied to the artificial neural network model in this study.The average relative error was only 3.81%,which can achieve the ideal prediction effect and provide a new way to effectively predict the strength of alkali-activated slag concrete.

关 键 词：碱激发矿渣混凝土 人工神经网络 强度 PCA-BP网络模型 

分 类 号：TU528.01[建筑科学—建筑技术科学]