Performance optimization of rigid pavement concrete using metakaolin treated RCA and silica fume with an experimental and machine learning based approach

采用实验和机器学习方法优化掺入偏高岭土处理再生混凝土骨料和硅粉的刚性路面混凝土的性能

阅读:1

Abstract

Concrete production has a drastic effect on the environment with ordinary portland cement (OPC) contributing approximately 6-8% to the world CO(2) emissions. The best solutions to reduce these effects is the use of recycled concrete aggregate (RCA) and secondary cementitious materials (SCM) as an alternative to natural aggregates and OPC. Nevertheless, RCA based on low-strength parent concrete is normally characterized by high porosity, low-bond mortar and low mechanical performance which restricts the scope of its structural use. This paper examines the improvement of RCA-based concrete using metakaolin (MK) slurry treatment and addition of silica fume (SF) at different dosages (2.5-10%) with the replacement contents of RCA being 0, 50, 75, and 100%. It has been experimentally found that the addition of MK and SF can significantly enhance mechanical strength and durability and adequately address the intrinsic weaknesses of low-grade RCA. The statistical validation with one-way ANOVA showed that all the P-values were less than 0.05 and proved that the improvements due to the addition of SCM and the adjustment of RCA were significant. In addition, 96 experimental and 48 literature-based datasets were used to predict and optimize compressive strength with the use of machine learning (ML) models. K-fold cross-validation was used to fine-tune hyperparameters and the Grey Wolf Optimizer (GWO) was used to optimize them. Extreme Gradient Boosting (XGB) gave the best accuracy with the highest R(2) of 0.949 (training) and 0.899 (testing) and low RMSE of 1.490 and 1.845 respectively. AdaBoost (ADB) also provided satisfactory results after XGB (R(2) = 0.929 training, 0.878 testing). In general, the findings substantiate the claim that the ensemble learning frameworks especially XGB are quite effective to derive complex relationships in RCA-based concrete data. RCA, SCMs (MK and SF) and predictive ML modeling can provide a sustainable mix design optimization route and structural life enhancement of RCA concrete in rigid pavement applications.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。