Predictive Modeling for Self-Compacting Concrete - Evaluating Machine Learning Approaches in Real-World Construction Scenarios

Aug 4, 2026

This study delivers a comprehensive review and analytical evaluation of machine learning methods for predicting self-compacting concrete (SCC) properties, integrating experimental data from existing literature to train and assess ML models. The work compares Artificial Neural Networks (ANN), Support Vector Machines (SVM), Decision Tree Regression (DTR) and related models, focusing on predictive accuracy under real-world construction scenarios. ANN demonstrates strong potential for complex data and adaptability but requires extensive high-quality datasets, while SVM offers robust generalisability with limited data and DTR variants balance accuracy and efficiency. Challenges such as data diversity, generalisability and practical applicability are discussed, with future research directions emphasizing hybrid models, larger datasets, and wider mixture-condition coverage to support efficient and sustainable SCC use.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages9

Reading Time23 min

AccessFree to download

Abstract

Keywords

1 Introduction

2 Methodology

ORCA – Online Research @ Cardiff This is an Open Access document downloaded from ORCA, Cardiff University's institutional repository:[https://orca.cardiff.ac.uk/id/eprint/182985/](https://orca.cardiff.ac.uk/id/eprint/182985/) This is the author’s version of a work that was submitted to / accepted for publication. Citation for final published version: Aldawish, Abdulaziz and Kulasegaram, Sivakumar 2024. Predictive modeling for self-compacting concrete: evaluating machine learning approaches in real-world construction scenarios. Presented at: 4th fib International Conference on Concrete Sustainability (ICCS2024), Guimarães, Portugal, 11–13 September 2024. Published in: Barros, Joaquim A. O., Cunha, Vítor M. C. F., Sousa, Hélder S., Matos, José C. and Sena-Cruz, José M. eds. 4th fib International Conference on Concrete Sustainability (ICCS2024) . Lecture Notes in Civil Engineering. Lecture Notes in Civil Engineering (573) Springer Nature Switzerland, pp. 236- 242. 10. 1007/978-3-031-80672-8_29 Publishers page: [https://doi.org/10.1007/978-3-031-80672-8_29](https://doi.org/10.1007/978-3-031-80672-8_29) Please note: Changes made as a result of publishing processes such as copy-editing, formatting and page numbers may not be reflected in this version. For the definitive version of this publication, please refer to the published source. You are advised to consult the publisher’s version if you wish to cite this paper. This version is being made available in accordance with publisher policies. See [http://orca.cf.ac.uk/policies.html](http://orca.cf.ac.uk/policies.html) for usage policies. Copyright and moral rights for publications made available in ORCA are retained by the copyright holders. Predictive Modeling for Self-Compacting Concrete: Evaluating Machine Learning Approaches in Real-World Construction Scenarios Abdulaziz Aldawish[0009-0007-0499-8815] and Sivakumar Kulasegaram12[0000-0002-9841-1339] 1* School of Engineering, Cardiff University, Cardiff, UK {Aldawisha,[kulasegarams}@cardiff.ac.uk](kulasegarams}@cardiff.ac.uk) Abstract. This study provides a comprehensive review and analysis of the applications of various machine learning techniques in predicting properties of self-compacting concrete (SCC) . This study also integrated experimental data from existing literature to build and evaluate ML models. We critically assess methodologies, strengthsand limitations of Artificial Neural Networks (ANN), Support Vector Machines (SVM), Decision Tree Regression s(DTR), and other machine learning models, emphasizing their predictive accuracy in real-world scenarios. We found that ANN showed significant promise for handling complex data structures and adaptability, but was dependent on extensive and high-quality datasets. SVM exceled in generalisability and effectiveness, even with limited data, while DTR and its advanced forms, such as XGBoost, offered a balance of accuracy and efficiency. The objective of this study was to identify the most effective model based on predictive accuracy and efficiency in realworld construction scenarios. Furthermore, this study explored challenges such as data diversity, model generalizability and real-world applicability. Future research should focus on hybrid models, expanding datasets, and applying these models to diverse concrete mixtures and conditions, offering significant implications for efficient and sustainable SCC use in construction. Keywords: Self-Compacting Concrete (SCC), machine learning (ML), of Artificial Neural Networks (ANN), Support Vector Machines (SVM), Decision Tree Regression (DTR) 1 Introduction Self-compacting concrete (SCC), conceptualized in 1986 by Professor Okamura[1], marked a significant advancement in construction materials; its unique ability to flow and self-compact without mechanical vibration revolutionized construction practices . Okamura's work led to the development of SCC with high flowability, with Ozawa creating the first

FAQs

Which machine learning models are assessed for predicting SCC properties?

The study critically assesses Artificial Neural Networks (ANN), Support Vector Machines (SVM), Decision Tree Regression (DTR) and other ML approaches including advanced tree-based forms such as XGBoost.

How does the paper integrate data to build and evaluate the ML models?

It integrates experimental data from existing literature to build and evaluate the ML models, using datasets gathered from published articles.

What key challenges are discussed for real-world SCC predictive modeling?

The paper highlights challenges including data diversity, model generalisability, and real-world applicability when deploying models for construction scenarios.

Aug 4, 2026

Upload by Theodora