基于堆叠融合技术的碱激发胶凝材料碳化性能预测  

Prediction of Carbonation Performance of Alkali-Activated Cementitious Materials Based on Stacked Ensemble Techniques

作　　者：谭绍恩 马倩敏 李平 TAN Shaoen;MA Qianmin;LI Ping(Faculty of Civil Engineering and Mechanics,Kunming University of Science and Technology,Kunming 650500,China;Yunnan Key Laboratory of Disaster Reduction in Civil Engineering,Kunming 650500,China;International Joint Laboratory for Green Construction and Intelligent Maintenance of Yunnan Province,Kunming 650500,China)

机构地区：[1]昆明理工大学建筑工程学院,昆明650500 [2]云南省土木工程防灾重点实验室,昆明650500 [3]云南省绿色建造与智慧运维国际联合实验室,昆明650500

出　　处：《材料导报》2024年第S02期288-296,共9页Materials Reports

基　　金：国家自然科学基金(52068038);云南省科技厅科技计划项目(202101AT070089,202202AG050020)。

摘　　要：碱激发胶凝材料碳化性能预测研究工作可对推进建筑业新质生产力和绿色低碳发展产生积极影响。本研究采用堆叠融合技术,集成随机森林、径向基函数和支持向量机三种模型对碱激发胶凝材料碳化性能进行了机器学习和预测研究。结果表明,基于现有数据集,三种基础模型已经展现出了良好的预测能力和稳定性。引入堆叠融合技术后,堆叠模型的均方误差(MSE)可控制在2%左右,拟合系数(R^(2))评分突破0.90,这表明堆叠融合模型表现出卓越的预测准确性、出色的泛化能力和稳健性。The prediction of carbonation performance of alkali-activated cementitious materials has a positive impact on promoting new productivity and green,low-carbon development in the construction industry.This study employs stacked ensemble techniques to integrate three models-random forest,radial basis function,and support vector machine-to conduct machine learning and prediction research on the carbonation performance of alkali-activated cementitious materials.The results show that each of the three base models has already demonstrated good predictive ability and stability based on the existing dataset.After introducing stacked ensemble techniques,the mean squared error(MSE)of the ensemble model is controlled at around 2%,and the R^(2)score exceeds 0.90.This indicates that the stacked ensemble model exhibits excellent predictive accuracy,generalization ability,and robustness.

关 键 词：碱激发胶凝材料 碳化 机器学习 堆叠融合模型 集成预测 预测准确性 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TU528[自动化与计算机技术—控制科学与工程]