Comparative analysis of flexural strength prediction in SFRC using frequentist, Bayesian, and Machine Learning approaches
Steel fiber reinforcement significantly enhances the flexural strength of concrete, which is vital for structural integrity. Annex L of the new Eurocode 2 classifies steel fiber-reinforced concrete by its flexural performance, aiding engineers in designing resilient structures. This study investigat...
| Autores: | , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Fecha de publicación: | 2024 |
| País: | España |
| Institución: | Universidad Rey Juan Carlos |
| Repositorio: | BURJC-Digital. Repositorio Institucional de la Universidad Rey Juan Carlos |
| OAI Identifier: | oai:burjcdigital.urjc.es:10115/40178 |
| Acceso en línea: | https://hdl.handle.net/10115/40178 |
| Access Level: | acceso abierto |
| Palabra clave: | Steel-fiber reinforced concrete Flexural behaviour Data-driven analysis Frequentist inference Bayesian inference Machine Learning |
| Sumario: | Steel fiber reinforcement significantly enhances the flexural strength of concrete, which is vital for structural integrity. Annex L of the new Eurocode 2 classifies steel fiber-reinforced concrete by its flexural performance, aiding engineers in designing resilient structures. This study investigates the flexural behavior of steel fiber-reinforced concrete (SFRC) using three data-driven methodologies: Frequentist Inference (FI), Bayesian Inference (BI), and Machine Learning (ML). A comprehensive database was constructed from three-point bending tests on SFRC specimens, encompassing various compressive strengths, fiber quantities, and geometric parameters, to identify key factors influencing material properties. The findings indicate that all three methodologies yield comparable predictive capabilities for flexural responses in SFRC. Notably, FI models emphasize the importance of compressive strength and fiber volume fraction, along with fiber properties such as non-dimensional length and tensile strength. BI models enhance predictive stability by integrating prior knowledge and quantifying uncertainty, demonstrating their advantage, particularly in data-scarce situations. Additionally, ML analysis reveals that linear regression (LR) models can achieve accuracy similar to or greater than that of more complex models. This research provides novel insights into the application of BI and ML in concrete technology, emphasizing their potential to enhance predictive modeling. Additionally, it offers practical guidelines for optimizing SFRC design through a case study that compares residual flexural strengths obtained via Bayesian analysis, classifying the material in accordance with Annex L of the new Eurocode 2 |
|---|