Evaluation of prediction model and machine learning methods on the shrinkage of self-compacting concrete
Journal
Magazine of Concrete Research
Journal Volume
78
Journal Issue
1-2
Start Page
1
End Page
17
ISSN
0024-9831
1751-763X
Date Issued
2025-12-09
Author(s)
Abstract
Shrinkage strain directly reduces the concrete volume and significantly affects the long-term durability of structures, particularly in self-compacting concrete (SCC). However, most existing models are not sufficiently accurate for predicting SCC shrinkage. To address this, a new SCC prediction model, called fib-SCC, was developed by calibrating the fib Model Code 2010 (MC2010) using an expanded laboratory database, comprising 1328 datasets for total shrinkage and 282 datasets for autogenous shrinkage, collected from a wide range of published studies. The fib-SCC model was evaluated through targeted SCC experiments based on varying fly ash contents (20–55% by weight of cement). Additionally, new prediction methods for SCC shrinkage behaviour are proposed using machine learning (ML) algorithms, including decision trees, the random forest, extra trees, gradient boosting (GB) and extreme gradient boosting (XGBoost). The results showed that the fib-SCC model had the highest accuracy among traditional models and other SCC prediction models. Nevertheless, the fib-SCC model was outperformed by the ML approaches, with GB and XGBoost demonstrating the highest accuracy in predicting total shrinkage.
Subjects
machine learning (ML)
non-destructive testing
plain concrete
prediction models
self-compacting concrete (SCC)
shrinkage prediction
Publisher
Emerald
Type
journal article