Work

High-temperature damage characterization and strength prediction of Ultra-High performance emery concrete: Combining acoustic emission and graph neural network

Year: 2025

Type: article

Abstract: Concrete structures exposed to high temperatures face significant safety challenges. This study systematically investigated the mechanical properties and damage behavior of Ultra-High performance emery concrete (UHPEC) after exposure to high temperatures. The surface morphology, mass loss rate, and residual compressive strength of UHPEC at various temperatures were studied. Acoustic emission (AE) analysis was used to evaluate UHPEC damage behavior. Material, temperature, and AE features were screened through Extreme Gradient Boosting (XGBoost) and correlation analysis, after which a graph sample and aggregate (GraphSAGE) model was employed to predict the residual compressive strength at elevated temperatures. The results indicate that UHPEC containing 50 % emery and 0.2 % polypropylene (PP) fibers showed the best thermal performance, retaining 127.6 % of its compressive strength after 400 °C and exhibiting no spalling at any test temperature. AE analysis reveals that crack propagation intensifies with rising exposure temperatures but decreases with higher emery replacement rates, indicating emery effectively suppresses high-temperature cracking. PP fibers also suppress cracking below 300°C; however, above 300°C, fully melted PP fibers form microchannels that enhance crack connectivity and propagation. XGBoost importance scores reveal that AE features were critical in predicting residual compressive strength at high temperatures. The GraphSAGE model achieved a test set coefficient of determination of 0.9593 and a mean absolute percentage error of 0.0517, demonstrating robust generalization and high predictive accuracy. • Physical properties and residual compressive strength of UHPEC were studied at various temperatures. • Damage evolution of UHPEC at 25-300°C was analyzed using AE parameters. • Conventional and AE features selected by XGBoost and correlation analysis. • GraphSAGE-based model developed for high-temperature strength prediction.

Source: Case Studies in Construction Materials

DOI: 10.1016/j.cscm.2025.e05502

Authors: Wei Du, Zixian Liu, Hua Yang, Yaojia Chen, Yubo Jiao

Institution: Beijing University of Technology

Language: English

Topic: Fire effects on concrete materials

Subfield: Civil and Structural Engineering

Field: Engineering

Domain: Physical Sciences

Open Access status: gold

Estimated APC paid: $2,660

Funder: National Natural Science Foundation of China