Efficient prediction of the load-carrying capacity of ECC-strengthened RC beams - An extra-gradient boosting machine learning method

As a result of the excellent crack width control capabilities of engineered cementitious composites (ECC), an ECC layer located in tension zones in reinforced concrete structures can transform wide harmful cracks into harmless dense microcracks. As a result, reinforced concrete structures are more durable and less prone to corrosion in aggressive environments. In the literature, there are equations that can predict the flexural capacity of ECCstrengthened RC beams. However, these equations are primarily regression-based and contain a limited number of parameters. With the help of the machine learning (ML) technique, a more comprehensive equation considering all the governing parameters can be proposed for ECC-strengthened RC beams. In the present study, a large database containing data from 217 ECC-strengthened beam specimens was collected from around two dozen publications. These data were then employed to develop an ML model using the XG boost algorithm. The model uses 20 input variables to predict the load-carrying capacity of the ECC-strengthened beam. The optimal model can produce a predicted-target dataset for the test dataset with an accuracy level of above 80%. Model parameters with the greatest significance according to Gini indexes were yield strength of the steel bars in the concrete substrate, beam depth, longitudinal reinforcement area, and ECC thickness. The analysis of SHAP (an acronym for SHapley Additive Explanations) values revealed a consistent pattern from the optimized model. Eventually, the study offers a free and easy-to-use graphical user interface to facilitate user interaction with the developed XG Boost model.