Improved static and impact properties of UHPFRC retrofitted with PU grout materials: Experiments and ML algorithms

• Published in: Results in engineering Vol. 23; p. 102655
• Main Authors: Al-shawafi, Ali, Zhu, Han, Laqsum, Saleh Ahmed, Haruna, S.I., Ibrahim, Yasser E.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.09.2024; Elsevier
• Subjects: And impact resistance; Drop-weight impact test; Machine learning; Polyurethane grout; UHPFRC; And impact resistance; Polyurethane grout; UHPFRC; Drop-weight impact test; Machine learning
• ISSN: 2590-1230; 2590-1230
• DOI: 10.1016/j.rineng.2024.102655

This study addresses the inherent brittleness of ultra-high-performance fiber re-inforced concrete (UHPFRC) by introducing a U-shaped specimen and investigating the impact strength of U-shaped UHPFRC retrofitted with polyurethane (PU) grout overlays. A drop-weight impact test was conducted on the U-shaped specimens utilizing a 2.1 kg weight. Various PU overlay thicknesses (5 mm, 10 mm, and 15 mm) were applied to the specimens. Machine learning techniques, specifically artificial neural networks (ANN) and support vector regression (SVR) were utilized to analyze the experimental data. Results indicate that UHPFRC cast with PU grout overlaid exhibit decrease in flexural strength compare to reference specimens. On the other hand, significance improvement in impact resistance were observed, with overlaid thickness. The addition of 5 mm, 10 mm, and 15 mm PU grout layers substantially improved the first crack strength of UHPFRC-5PU, UHPFRC-10PU, and UHPFRC-15PU specimens by 94 %, 340.3 %, and 600 %, respectively, compared to UHPFRC-0PU. Machine learning models accurately predicted failure crack strength (N2), with ANN and SVR achieving determination correlation (R2) values of 0.9838 and 0.9816 during the training and testing phases, respectively. •The synergic effects of steel fiber and PU grout overlaid on the UHPC was investigated.•Failure crack strength was estimated using machine learning.•Impact strength was significantly improved due to the addition of PUG overlaid.•ANN and SVR models predict the failure strength with high accuracy.