Axial Capacity Estimation of FRP-strengthened Corroded Concrete Columns

Abstract

This research presents a machine learning (ML) based model to estimate the axial strength of corroded RC columns reinforced with fiber-reinforced polymer (FRP) composites. Estimating the axial strength of corroded columns is complex due to the intricate interplay between corrosion and FRP reinforcement. To address this, a dataset of 102 samples from various literature sources was compiled. Subsequently, this dataset was employed to create and train the ML models. The parameters influencing axial strength included the geometry of the column, properties of the FRP material, degree of corrosion, and properties of the concrete. Considering the scarcity of reliable design guidelines for estimating the axial strength of RC columns considering corrosion effects, artificial neural network (ANN), Gaussian process regression (GPR), and support vector machine (SVM) techniques were employed. These techniques were used to predict the axial strength of corroded RC columns reinforced with FRP. When comparing the results of the proposed ML models with existing design guidelines, the ANN model demonstrated higher predictive accuracy. The ANN model achieved an R-value of 98.08% and an RMSE value of 132.69 kN which is the lowest among all other models. This model fills the existing gap in knowledge and provides a precise means of assessment. This model can be used in the scientific community by researchers and practitioners to predict the axial strength of FRP-strengthened corroded columns. In addition, the GPR and SVM models obtained an accuracy of 98.26% and 97.99%, respectively.

Keywords

• rc column
• corrosion
• axial capacity
• machine learning frp
• ann