Metaheuristic optimization of machine learning models for strength prediction of high-performance self-compacting alkali-activated slag concrete

The present study focuses on producing high-performance eco-efficient alternatives to conventional cement-based composites. The study is divided into two parts. The first part comprises of production of high-strength self-compacting alkali-activated slag concrete (SC-AASC) with GGBFS as a primary bi...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Multiscale and Multidisciplinary Modeling, Experiments and Design Jg. 7; H. 3; S. 2901 - 2928
Hauptverfasser: Parhi, Suraj Kumar, Panda, Soumyaranjan, Dwibedy, Saswat, Panigrahi, Saubhagya Kumar
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Cham Springer International Publishing 01.07.2024
Schlagworte:
Characterization and Evaluation of Materials
Engineering
Mathematical Applications in the Physical Sciences
Mechanical Engineering
Numerical and Computational Physics
Original Paper
Simulation
Solid Mechanics
High-strength
Self-compacting
Sensitivity analysis
GGBFS
XGBoost
Machine learning
Alkali-activated concrete
SHO
ISSN:2520-8160, 2520-8179
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The present study focuses on producing high-performance eco-efficient alternatives to conventional cement-based composites. The study is divided into two parts. The first part comprises of production of high-strength self-compacting alkali-activated slag concrete (SC-AASC) with GGBFS as a primary binder. The second part deals with the development of a prediction model to estimate the mechanical strength of developed concrete. In this study, to achieve high-performance SC-AASC, the alkali activator solution content varied from 220 to 190 kg/m 3 , and the AAS/binder ratio varied between 0.47 and 0.36. The SP percentage fluctuated between 6 and 7%, while the additional water percentage was maintained between 21 and 24%. The approach used to obtain the high-performance SC-AASC was found to be competent as all the mix resulted in satisfactory performance for both fresh and hardened properties. For M45 graded SC-AASC, using 200 kg/m 3 of AAS with an AAS/binder ratio of 0.39 resulted in higher strength, while for M60 grade, 190 kg/m 3 of AAS with an AAS/binder ratio of 0.36 yielded stronger concrete. Additionally, a 6% SP and 24% extra water content enhanced workability for both M45 and M60 grade SC-AASC. A database of 135 observations was developed from the experimental study. The compressive strength and split tensile strength of SC-AASC were predicted using six machine-learning algorithms. The hyperparameters of all the models were optimized using the metaheuristic spotted hyena optimization technique. Optimized XGBoost outperformed other models scoring a higher R 2 of 0.97 and lower value of error parameters on both datasets. A comparison was drawn with previously published models to check the efficacy of the developed model. The Sobol and FAST global sensitivity analysis resulted in the AAS/binder ratio, AAS content, GGBFS content, and Curing days being most influential regarding the strength of SC-AASC.
ISSN:2520-8160
2520-8179
DOI:10.1007/s41939-023-00349-4