Artificial intelligence-based optimized models for predicting the slump and compressive strength of sustainable alkali-derived concrete

•AI tools (GEP and MEP) were used for developing prediction models for AAC properties.•The sensitivity analysis of the database showed the relevance of input parameters.•The resulting prediction models based on empirical equations agreed well with targets.•MEP models yielded more accurate estimates...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Construction & building materials Jg. 409; S. 134092
Hauptverfasser: Zou, Baoping, Wang, Yanbing, Nasir Amin, Muhammad, Iftikhar, Bawar, Khan, Kaffayatullah, Ali, Mujahid, Althoey, Fadi
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 15.12.2023
Schlagworte:
Alkali-activated concrete
Compressive strength
Gene expression programming
Multi-expression programming
Sensitivity analysis
Slump
Taylor diagram
Sensitivity analysis
Multi-expression programming
Alkali-activated concrete
Compressive strength
Slump
Taylor diagram
Gene expression programming
ISSN:0950-0618
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:•AI tools (GEP and MEP) were used for developing prediction models for AAC properties.•The sensitivity analysis of the database showed the relevance of input parameters.•The resulting prediction models based on empirical equations agreed well with targets.•MEP models yielded more accurate estimates than GEP based on R2 and statistical measures.•The prediction models may estimate the AAC properties with diverse input parameters. Alkali-activated materials (AAMs) are a potential class of construction materials that are well-known for their versatility and capacity for long-term sustainability. As a result of its ability to lessen the negative effects that the building industry has on the environment, AAMs have become increasingly popular in recent years. However, it can be difficult and time-consuming to figure out what proportions of alkali-activated concrete (AAC) would work best for a given project. Compressive strength (CS) and slump, both of which are important properties of AAC's viability in construction, were predicted using machine learning (ML) techniques, such as multi-expression programming (MEP)and gene expression programming (GEP)in this study. The mathematical formulations of AAC for both slump and CS for theAAC were effectively derived with the application of these MLapproaches. According to the study's findings, MEP models performed better than GEP models in making accurate predictions, with MEP achieving R2 values of 0.92 and 0.93 for slump and CS in AAC, respectively, whereas GEP provided R2 values of 0.86 and 0.89. The hyper-parameters of the AI models were fine-tuned, and the models were verified with statistical measurements and Taylor diagrams. It's possible that using the findings from sensitivity analysis to estimate the relative importance of factors impacting the slumpand CS of AAC might be helpful. The artificial intelligence-based models that were built showed a strong connection with the desired outcomes, suggesting that they might be used to estimate the slump and CS of AAC for different values of the input components.
ISSN:0950-0618
DOI:10.1016/j.conbuildmat.2023.134092