Dynamics of Compressive Performance in Textile-Reinforced Concrete: A Machine-Learning Approach for Failure Mode and Strength Prediction

Abstract This study presents a computational failure framework that incorporates decision tree classification and machine-learning modeling to identify failure modes in textile-reinforced concrete columns and predict their strength. To achieve this, a decision tree classifier was developed using an experimentally derived dataset that considered geometry information, matrix parameters, fiber type (basalt, carbon, steel, glass, PBO), and concrete core properties to detect tensile-jacket rupture and interfacial debonding failure modes. A machine-learning model was subsequently proposed to estimate the ultimate confined compressive strength, using the unconfined concrete strength, confinement ratio, fiber tensile capacity, and matrix characteristics. The decision tree achieved an acceptable classification accuracy, revealing that the fiber ratio and concrete strength were the most important parameters for predicting the failure mode. Moreover, the developed explicit predictive expression for the design applications of the ultimate strength of columns demonstrated a suitable correlation with experimental strength measurements. The proposed methodology not only elucidates the key factors affecting failure behavior but also provides practical tools for forensic investigation and reliability-based design of fiber-reinforced confinement systems.