基于机器学习的煤矸石粉混凝土抗压强度预测

打开文本图片集

中图分类号：TU528 文献标志码：A 文章编号：1005-8249（2025）04-0157-04

DOI：10.19860/j.cnki.issn1005-8249.2025.04.028

ZHAO Qi， LAN Yongqiang

（Department of Infrastructure，Yan’an University，Yan’an 716OOO，China）

Abstract：To improve theaccuracyof predicting thecompressivestrength ofcoalgangue powderconcrete，adatasetconsisting of 70 groups of compressive strength data for coal gangue powder concrete was constructed.Three classic machine learning ensemblealgorithms were introduced，and three prediction modelsforthecompresive strengthofcoal gangue powderconcrete considering multiplefactors wereestablished.Theimpactofdiferent featurecombinationsonmodelperformancewasstudied. Theresultsshowthatthe XGBostalgorithm modelistheoptimal modelforpredictingthecompressve strengthofcoalgangue powder concrete，with a coefficient of determination（ R2 ）of O.857 and a root mean square error of 2.O6；Cement dosage and curing age are thetwo mostimportant feature indicators；whenusing 6features including water-binderratio，cement dosage， coal ganguepowderdosage，sand dosage，graveldosage，and curing age，the average coeficientof determinationof thethree diferentalgorithmsreachesthehighestvalueofO.844，whichreflects thehighaccuracyof themodel.Theresearchresultscan provide a fast and reliable decision-making basis for the engineering application of coal gangue powder concrete.

Key words： coal gangue powder； concrete; compressive strength； forecast; machine learning

0 引言

煤矸石粉应用于混凝土中，是实现煤矸石“变废为宝”的重要处置方式[1]。（剩余4997字）