Unbiased Model of the Contribution of NSM CFRP Systems to Shear Resistance in RC Beams Utilizing Multiobjective Optimization

This study presents the development of a predictive model for assessing the contribution of near-surface mounted (NSM) carbon fiber–reinforced polymer (CFRP) systems to the shear resistance of RC beams. The primary motivation is to address the limitations of existing models, which often exhibit limited accuracy. The study utilizes an up-to-date data set comprising 195 RC beams derived from experimental results, which underwent careful outlier elimination using the isolation forest method to retain only 114 relevant data points for analysis. By integrating truss analogy, the simplified modified compression field theory, and multiobjective optimization techniques within a double Loop iteration procedure, a semiempirical model was derived to predict the shear resistance of RC beams shear-strengthened with NSM CFRP reinforcements. The optimization process utilized Nondominated Sorting Genetic Algorithm II (NSGA-II) to execute Pareto optimization, effectively balancing competing objectives. By applying NSGA-II, Pareto-optimal solutions were identified, and a unique solution was selected based on the shortest Euclidean distance to the ideal point among the Pareto front solutions. This approach ensured that the model accurately captures the trade-offs between different performance criteria, resulting in a predictive model of superior accuracy and robustness. The predictive model was validated against the collected data set and demonstrated superior accuracy compared with existing models.