Strength Prediction of Sustainable UHPC Via VAE-Augmented Data and Explainable Boosting Models

A. I.B. Farouk; Salah U. Al-Dulaijan; Mohammed A. Al-Osta; S. I. Haruna; Yasser E. Ibrahim; Suleiman Abdulrahman

doi:10.1007/978-981-92-1595-9_33

Strength Prediction of Sustainable UHPC Via VAE-Augmented Data and Explainable Boosting Models

A. I.B. Farouk^*
, Salah U. Al-Dulaijan
, Mohammed A. Al-Osta
, S. I. Haruna
, Yasser E. Ibrahim
, Suleiman Abdulrahman

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

This study develops an interpretable data-centric machine learning paradigm to predict the compressive strength of ultra-high-performance concrete (UHPC) with recycled concrete powder (RCP). Given limited experimental data, an experimental dataset of 80 samples was augmented with 300 synthesized concrete mix designs to improve the model. We evaluated the performance of two machine learning methods using extreme learning machine (ELM) and CatBoost, employing three different data strategies: experimental only (M1), synthetic only (M2), and combined (M3). Our results showed that CatBoost in the combined strategy (M3) outperformed the competition with an R² of 0.968 and the best error metrics in the test phase. To demonstrate the feature contributions, we used SHAP (SHapley Additive exPlanations) and found that the most important features explaining compressive strength included cement, steel fibers, and water. Our proposed method demonstrates that combining generative data with explainable machine learning significantly improves both predictive performance and transparency, suggesting potential for designing and optimizing concrete mix recipes for sustainable UHPC.

Original language	English
Title of host publication	Proceedings of the 3rd International Conference on Sustainability
Subtitle of host publication	Developments and Innovations Vol 1 - Proceedings of ICSDI 2026 Vol 1
Editors	Mohamed Ezzat Al-Atroush, Zahiraniza Mustaffa, Moustafa M. Nasralla, Abdulhakim Almajid
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	317-324
Number of pages	8
ISBN (Print)	9789819215942
DOIs	https://doi.org/10.1007/978-981-92-1595-9_33
State	Published - 2026
Event	3rd International Conference on Sustainability: Developments and Innovations, ICSDI 2026 - Riyadh, Saudi Arabia Duration: 8 Feb 2026 → 12 Feb 2026

Publication series

Name	Lecture Notes in Civil Engineering
Volume	876 LNCE
ISSN (Print)	2366-2557
ISSN (Electronic)	2366-2565

Conference

Conference	3rd International Conference on Sustainability: Developments and Innovations, ICSDI 2026
Country/Territory	Saudi Arabia
City	Riyadh
Period	8/02/26 → 12/02/26

Bibliographical note

Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.

Keywords

Compressive strength prediction
Recycled concrete powder (RCP)
SHAP analysis
Ultra-high-performance concrete (UHPC)
Variational autoencoder (VAE)

ASJC Scopus subject areas

Civil and Structural Engineering

Access to Document

10.1007/978-981-92-1595-9_33

Cite this

Farouk, A. I. B., Al-Dulaijan, S. U., Al-Osta, M. A., Haruna, S. I., Ibrahim, Y. E., & Abdulrahman, S. (2026). Strength Prediction of Sustainable UHPC Via VAE-Augmented Data and Explainable Boosting Models. In M. E. Al-Atroush, Z. Mustaffa, M. M. Nasralla, & A. Almajid (Eds.), Proceedings of the 3rd International Conference on Sustainability: Developments and Innovations Vol 1 - Proceedings of ICSDI 2026 Vol 1 (pp. 317-324). (Lecture Notes in Civil Engineering; Vol. 876 LNCE). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-981-92-1595-9_33

Farouk, A. I.B. ; Al-Dulaijan, Salah U. ; Al-Osta, Mohammed A. et al. / Strength Prediction of Sustainable UHPC Via VAE-Augmented Data and Explainable Boosting Models. Proceedings of the 3rd International Conference on Sustainability: Developments and Innovations Vol 1 - Proceedings of ICSDI 2026 Vol 1. editor / Mohamed Ezzat Al-Atroush ; Zahiraniza Mustaffa ; Moustafa M. Nasralla ; Abdulhakim Almajid. Springer Science and Business Media Deutschland GmbH, 2026. pp. 317-324 (Lecture Notes in Civil Engineering).

@inproceedings{3d18b147b31c4a5094364d78d792e0ad,

title = "Strength Prediction of Sustainable UHPC Via VAE-Augmented Data and Explainable Boosting Models",

abstract = "This study develops an interpretable data-centric machine learning paradigm to predict the compressive strength of ultra-high-performance concrete (UHPC) with recycled concrete powder (RCP). Given limited experimental data, an experimental dataset of 80 samples was augmented with 300 synthesized concrete mix designs to improve the model. We evaluated the performance of two machine learning methods using extreme learning machine (ELM) and CatBoost, employing three different data strategies: experimental only (M1), synthetic only (M2), and combined (M3). Our results showed that CatBoost in the combined strategy (M3) outperformed the competition with an R2 of 0.968 and the best error metrics in the test phase. To demonstrate the feature contributions, we used SHAP (SHapley Additive exPlanations) and found that the most important features explaining compressive strength included cement, steel fibers, and water. Our proposed method demonstrates that combining generative data with explainable machine learning significantly improves both predictive performance and transparency, suggesting potential for designing and optimizing concrete mix recipes for sustainable UHPC.",

keywords = "Compressive strength prediction, Recycled concrete powder (RCP), SHAP analysis, Ultra-high-performance concrete (UHPC), Variational autoencoder (VAE)",

author = "Farouk, \{A. I.B.\} and Al-Dulaijan, \{Salah U.\} and Al-Osta, \{Mohammed A.\} and Haruna, \{S. I.\} and Ibrahim, \{Yasser E.\} and Suleiman Abdulrahman",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.; 3rd International Conference on Sustainability: Developments and Innovations, ICSDI 2026 ; Conference date: 08-02-2026 Through 12-02-2026",

year = "2026",

doi = "10.1007/978-981-92-1595-9\_33",

language = "English",

isbn = "9789819215942",

series = "Lecture Notes in Civil Engineering",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "317--324",

editor = "Al-Atroush, \{Mohamed Ezzat\} and Zahiraniza Mustaffa and Nasralla, \{Moustafa M.\} and Abdulhakim Almajid",

booktitle = "Proceedings of the 3rd International Conference on Sustainability",

address = "Germany",

}

Farouk, AIB , Al-Dulaijan, SU , Al-Osta, MA, Haruna, SI, Ibrahim, YE & Abdulrahman, S 2026, Strength Prediction of Sustainable UHPC Via VAE-Augmented Data and Explainable Boosting Models. in ME Al-Atroush, Z Mustaffa, MM Nasralla & A Almajid (eds), Proceedings of the 3rd International Conference on Sustainability: Developments and Innovations Vol 1 - Proceedings of ICSDI 2026 Vol 1. Lecture Notes in Civil Engineering, vol. 876 LNCE, Springer Science and Business Media Deutschland GmbH, pp. 317-324, 3rd International Conference on Sustainability: Developments and Innovations, ICSDI 2026, Riyadh, Saudi Arabia, 8/02/26. https://doi.org/10.1007/978-981-92-1595-9_33

Strength Prediction of Sustainable UHPC Via VAE-Augmented Data and Explainable Boosting Models. / Farouk, A. I.B.; Al-Dulaijan, Salah U.; Al-Osta, Mohammed A. et al.
Proceedings of the 3rd International Conference on Sustainability: Developments and Innovations Vol 1 - Proceedings of ICSDI 2026 Vol 1. ed. / Mohamed Ezzat Al-Atroush; Zahiraniza Mustaffa; Moustafa M. Nasralla; Abdulhakim Almajid. Springer Science and Business Media Deutschland GmbH, 2026. p. 317-324 (Lecture Notes in Civil Engineering; Vol. 876 LNCE).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Strength Prediction of Sustainable UHPC Via VAE-Augmented Data and Explainable Boosting Models

AU - Farouk, A. I.B.

AU - Al-Dulaijan, Salah U.

AU - Al-Osta, Mohammed A.

AU - Haruna, S. I.

AU - Ibrahim, Yasser E.

AU - Abdulrahman, Suleiman

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.

PY - 2026

Y1 - 2026

N2 - This study develops an interpretable data-centric machine learning paradigm to predict the compressive strength of ultra-high-performance concrete (UHPC) with recycled concrete powder (RCP). Given limited experimental data, an experimental dataset of 80 samples was augmented with 300 synthesized concrete mix designs to improve the model. We evaluated the performance of two machine learning methods using extreme learning machine (ELM) and CatBoost, employing three different data strategies: experimental only (M1), synthetic only (M2), and combined (M3). Our results showed that CatBoost in the combined strategy (M3) outperformed the competition with an R2 of 0.968 and the best error metrics in the test phase. To demonstrate the feature contributions, we used SHAP (SHapley Additive exPlanations) and found that the most important features explaining compressive strength included cement, steel fibers, and water. Our proposed method demonstrates that combining generative data with explainable machine learning significantly improves both predictive performance and transparency, suggesting potential for designing and optimizing concrete mix recipes for sustainable UHPC.

AB - This study develops an interpretable data-centric machine learning paradigm to predict the compressive strength of ultra-high-performance concrete (UHPC) with recycled concrete powder (RCP). Given limited experimental data, an experimental dataset of 80 samples was augmented with 300 synthesized concrete mix designs to improve the model. We evaluated the performance of two machine learning methods using extreme learning machine (ELM) and CatBoost, employing three different data strategies: experimental only (M1), synthetic only (M2), and combined (M3). Our results showed that CatBoost in the combined strategy (M3) outperformed the competition with an R2 of 0.968 and the best error metrics in the test phase. To demonstrate the feature contributions, we used SHAP (SHapley Additive exPlanations) and found that the most important features explaining compressive strength included cement, steel fibers, and water. Our proposed method demonstrates that combining generative data with explainable machine learning significantly improves both predictive performance and transparency, suggesting potential for designing and optimizing concrete mix recipes for sustainable UHPC.

KW - Compressive strength prediction

KW - Recycled concrete powder (RCP)

KW - SHAP analysis

KW - Ultra-high-performance concrete (UHPC)

KW - Variational autoencoder (VAE)

UR - https://www.scopus.com/pages/publications/105048867055

U2 - 10.1007/978-981-92-1595-9_33

DO - 10.1007/978-981-92-1595-9_33

M3 - Conference contribution

AN - SCOPUS:105048867055

SN - 9789819215942

T3 - Lecture Notes in Civil Engineering

SP - 317

EP - 324

BT - Proceedings of the 3rd International Conference on Sustainability

A2 - Al-Atroush, Mohamed Ezzat

A2 - Mustaffa, Zahiraniza

A2 - Nasralla, Moustafa M.

A2 - Almajid, Abdulhakim

PB - Springer Science and Business Media Deutschland GmbH

T2 - 3rd International Conference on Sustainability: Developments and Innovations, ICSDI 2026

Y2 - 8 February 2026 through 12 February 2026

ER -

Farouk AIB , Al-Dulaijan SU , Al-Osta MA, Haruna SI, Ibrahim YE, Abdulrahman S. Strength Prediction of Sustainable UHPC Via VAE-Augmented Data and Explainable Boosting Models. In Al-Atroush ME, Mustaffa Z, Nasralla MM, Almajid A, editors, Proceedings of the 3rd International Conference on Sustainability: Developments and Innovations Vol 1 - Proceedings of ICSDI 2026 Vol 1. Springer Science and Business Media Deutschland GmbH. 2026. p. 317-324. (Lecture Notes in Civil Engineering). doi: 10.1007/978-981-92-1595-9_33

Strength Prediction of Sustainable UHPC Via VAE-Augmented Data and Explainable Boosting Models

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this