Assessment of Axial-Flexural Behavior of Reinforced Concrete Column Wrapped with CFRP Using Probabilistic Machine Learning Model

Using Carbon fiber-reinforced polymer (CFRP) to enhance the capacity of structural reinforced-concrete members is a practical and fast approach that is increasingly recognized among the civil engineering community. However, in order to effectively and economically apply this approach, it is important to reliably estimate the CFRP-strengthened members’ performance based on their input parameters, such as member length, cross-section sizes, material properties, and external loadings. Toward this purpose, this study aims to develop a predictive model that can evaluate the axial-flexural behavior of reinforced concrete columns enhanced by CFRP based on, first, extensive columns data collected from the literature and experiments realized by the authors. The second key component of the proposed method is the probabilistic machine learning algorithm, namely, the quantile regression forest, which can account for the variance of input parameters among different experimental settings and provide the confidence intervals of predictions rather than some predicted values. Such types of results help engineers avoid under/over-estimate the members’ performance and produce adequate structural solutions.