Machine Learning Analysis for the Toughness Characteristics of Fiber ‎Reinforced Concrete

Abstract

The mechanical performance of fiber-reinforced concrete (FRC), especially its flexural strength and toughness measured by the load-deflection response, depends strongly on fiber characteristics such as volume fraction, length, tensile strength, shape, and type. While extensive research has explored these factors, much of it relies on idealized laboratory data, limiting practical applicability. This study addresses this gap by using real-world FRC data to develop predictive machine learning (ML) models that capture the combined influence of fiber and concrete properties on toughness. A comprehensive dataset of 146 FRC samples compiled from prior studies was analyzed. Four regression models—Random Forest (RF), Gradient Boosting (GB), Linear Regression, and Support Vector Regression—were trained and evaluated to predict the area under the load-deflection curve, a key indicator of toughness. The GB model achieved the best performance, with a coefficient of determination (R²) of 0.83 and a mean absolute error (MAE) of approximately 39, followed closely by RF (R² = 0.79). Feature importance analysis identified fiber volume fraction, fiber type and shape, and flexural strength as the most influential factors in enhancing toughness. This research provides a robust, data-driven predictive tool for estimating FRC toughness based on key physical and mechanical properties, offering valuable insights for engineers to optimize fiber-reinforced composite designs in practical structural applications.