Open Access

ARTICLE

High-Fidelity Machine Learning Framework for Fracture Energy Prediction in Fiber-Reinforced Concrete

Ala’a R. Al-Shamasneh¹, Faten Khalid Karim², Arsalan Mahmoodzadeh^3,*, Abdulaziz Alghamdi⁴, Abdullah Alqahtani⁵, Shtwai Alsubai⁵, Abed Alanazi⁵

1 Department of Computer Science, College of Computer & Information Sciences, Prince Sultan University, Rafha Street, Riyadh, 11586, Saudi Arabia
2 Department of Computer Sciences, College of Computer and Information Sciences, Princess Nourah Bint Abdulrahman University, P.O. Box 84428, Riyadh, 11671, Saudi Arabia
3 Center of Research and Strategic Studies, Lebanese French University, Erbil, 44001, Iraq
4 Department of Civil Engineering, Faculty of Engineering, University of Tabuk, Tabuk, 47512, Saudi Arabia
5 Department of Computer Science, College of Computer Engineering and Sciences in Al-Kharj, Prince Sattam bin Abdulaziz University, P.O. Box 151, Al-Kharj, 11942, Saudi Arabia

* Corresponding Author: Arsalan Mahmoodzadeh. Email:

Computer Modeling in Engineering & Sciences 2025, 144(2), 1573-1606. https://doi.org/10.32604/cmes.2025.068887

Received 09 June 2025; Accepted 07 August 2025; Issue published 31 August 2025

Abstract

The fracture energy of fiber-reinforced concrete (FRC) affects the durability and structural performance of concrete elements. Advancements in experimental studies have yet to overcome the challenges of estimating fracture energy, as the process remains time-intensive and costly. Therefore, machine learning techniques have emerged as powerful alternatives. This study aims to investigate the performance of machine learning techniques to predict the fracture energy of FRC. For this purpose, 500 data points, including 8 input parameters that affect the fracture energy of FRC, are collected from three-point bending tests and employed to train and evaluate the machine learning techniques. The findings showed that Gaussian process regression (GPR) outperforms all other models in terms of predictive accuracy, achieving the highest R² of 0.93 and the lowest RMSE of 13.91 during holdout cross-validation. It is then followed by support vector regression (SVR) and extreme gradient boosting regression (XGBR), whereas K-nearest neighbours (KNN) and random forest regression (RFR) show the weakest predictions. The superiority of GPR is further reinforced in a 5-fold cross-validation, where it consistently delivers an average R² above 0.96 and ranks highest in overall predictive performance. Empirical testing with additional sample sets validates GPR’s model on the key mix parameter’s impact on fracture energy, cementing its claim. The Fly-Ash cement exhibits the greatest fracture energy due to superior fiber-matrix interaction, whereas the glass fiber dominates energy absorption amongst the other types of fibers. In addition, increasing the water-to-cement (W/C) ratio from 0.30 to 0.50 yields a significant improvement in fracture energy, which aligns well with the machine learning predictions. Similarly, loading rate positively correlates with fracture energy, highlighting the strain-rate sensitivity of FRC. This work is the missing link to integrate experimental fracture mechanics and computational intelligence, optimally and reasonably predicting and refining the fracture energy of FRC.

---

Keywords

---