Modelling uniaxial compressive strength of lightweight self-compacting concrete using random forest regression

Junfei Zhang, Guowei Ma, Yimiao Huang, Junbo sun, Farhad Aslani, Brett Nener

Research output: Contribution to journalArticle

23 Citations (Scopus)

Abstract

Self-compacting concrete (SCC) can achieve compaction into every part of the formwork through its own weight without any segregation of the coarse aggregate. Lightweight concrete (LWC) can reduce the dead load of the structure by incorporating the lightweight aggregate (LWA). In recent years, more and more studies have focused on combining the advantages of SCC and LWC to produce lightweight self-compacting concrete (LWSCC). As one of the most important mechanical properties, uniaxial compressive strength (UCS) values need to be tested before field application of this new material. However, conducting UCS tests with multiple influencing variables is time-consuming and costly. To address this issue, this paper proposed, for the first time, a beetle antennae search (BAS) algorithm based random forest (RF) model to accurately and effectively predict the UCS of LWSCC. This model was developed and verified using data from LWSCC laboratory formulation. Results show that BAS was efficient in searching the optimum hyper-parameters of RF. The proposed BAS-RF model achieved high predictive accuracy indicated by a high correlation coefficient (0.97). In addition, by measuring the variable importance, we conclude that temperature was the most sensitive to UCS development, followed by scoria content and water-to-binder (w/b) ratio, while UCS was less sensitive to fiber content. This pioneering work provides a simple and convenient method for evaluating UCS of LWSCC at varying temperatures.

Original languageEnglish
Pages (from-to)713-719
Number of pages7
JournalConstruction and Building Materials
Volume210
DOIs
Publication statusPublished - 20 Jun 2019

Fingerprint Dive into the research topics of 'Modelling uniaxial compressive strength of lightweight self-compacting concrete using random forest regression'. Together they form a unique fingerprint.

Cite this