Publication
Data-driven analytical and ML-based models for shear-strengthened RC beams using FRP systems
| Summary: | Reinforced Concrete (RC) structures continue to dominate the construction industry due to their extensive use, durability and economic efficiency. However, as safety formats and performance demands evolve, many of these structures fail to comply with current requirements. Demolition, though sometimes necessary, is often suboptimal due to substantial economic costs and severe environmental consequences, such as waste generation and increased carbon emissions. Consequently, structural strengthening has emerged as a sustainable and cost-effective alternative. This thesis focuses on enhancing the shear resistance of RC beams using Carbon Fibre-Reinforced Polymer (CFRP) reinforcements, with a focus on two widely employed techniques: the Externally Bonded Reinforcement (EBR) and Near-Surface Mounted (NSM) techniques. Existing predictive models for these techniques exhibit insufficient accuracy, often resulting from oversimplified assumptions and failure to capture the complex, nonlinear interactions among various structural parameters. To address these limitations, this research introduces a suite of innovative models designed to significantly improve predictive accuracy and reliability. A comprehensive experimental database was compiled, and advanced data preprocessing methods, including outlier detection and data transformation, were implemented to improve model performance. Analytical models were developed with integrated calibration factors to minimize error variability and ensure consistent predictions across diverse beam characteristics. Additionally, a Reliability Analysis framework was established to calibrate resistance reduction factors, ensuring the proposed models align with safety and cost-effectiveness standards for practical design applications. Furthermore, Machine Learning models were developed, demonstrating substantial advancements in predictive accuracy and reliability compared to traditional analytical approaches. Overall, this research presents significant contributions to the field, offering improved design models that promote the sustainable and safe use of CFRP for shear strengthening, while laying the groundwork for future advancements. |
|---|---|
| Main Authors: | Mohammadi, Amir Hossein |
| Subject: | CFRP strengthened RC beam Design model for shear resistance Reliability analysis Resistance reduction factor Machine learning Viga de betão armado reforçada com CFRP Modelo de projeto para resistência ao corte Análise de confiabilidade Fator de redução da resistência Engenharia e Tecnologia::Engenharia Civil |
| Year: | 2025 |
| Country: | Portugal |
| Document type: | doctoral thesis |
| Access type: | open access |
| Associated institution: | Universidade do Minho |
| Language: | English |
| Origin: | RepositóriUM - Universidade do Minho |
| Summary: | Reinforced Concrete (RC) structures continue to dominate the construction industry due to their extensive use, durability and economic efficiency. However, as safety formats and performance demands evolve, many of these structures fail to comply with current requirements. Demolition, though sometimes necessary, is often suboptimal due to substantial economic costs and severe environmental consequences, such as waste generation and increased carbon emissions. Consequently, structural strengthening has emerged as a sustainable and cost-effective alternative. This thesis focuses on enhancing the shear resistance of RC beams using Carbon Fibre-Reinforced Polymer (CFRP) reinforcements, with a focus on two widely employed techniques: the Externally Bonded Reinforcement (EBR) and Near-Surface Mounted (NSM) techniques. Existing predictive models for these techniques exhibit insufficient accuracy, often resulting from oversimplified assumptions and failure to capture the complex, nonlinear interactions among various structural parameters. To address these limitations, this research introduces a suite of innovative models designed to significantly improve predictive accuracy and reliability. A comprehensive experimental database was compiled, and advanced data preprocessing methods, including outlier detection and data transformation, were implemented to improve model performance. Analytical models were developed with integrated calibration factors to minimize error variability and ensure consistent predictions across diverse beam characteristics. Additionally, a Reliability Analysis framework was established to calibrate resistance reduction factors, ensuring the proposed models align with safety and cost-effectiveness standards for practical design applications. Furthermore, Machine Learning models were developed, demonstrating substantial advancements in predictive accuracy and reliability compared to traditional analytical approaches. Overall, this research presents significant contributions to the field, offering improved design models that promote the sustainable and safe use of CFRP for shear strengthening, while laying the groundwork for future advancements. |
|---|