Experimental Investigation and Machine Learning-Based Prediction of the Mechanical and Durability Performance of Nano-Modified Fiber-Reinforced Concrete

Abstract

With the growing demand for high performance and durable concrete, advanced cementitious composites with nanomaterials and fiber reinforcement have been developed. Nano-modified fiber reinforced concrete can improve matrix densification, crack resistance, mechanical performance and durability through complementary mechanisms at different scales but its performance is dependent on complex interactions among binder composition, water to binder ratio, nanomaterial dosage, fiber content, admixture dosage, aggregate characteristics and curing age so conventional trial-and-error mix design is time consuming and resource intensive. In this paper an integrated experimental and AI based machine learning framework is proposed to investigate and predict the mechanical and durability performance of nano-modified fiber reinforced concrete. Nano silica is considered as the main nanomodifying material while fiber reinforcement is incorporated in controlled proportions to study its interaction with nano scale matrix modification. An experimental program is proposed to evaluate fresh properties, compressive strength, split tensile strength, flexural strength, water absorption, sorptivity and some durability characteristics at different curing ages. The experimentally generated dataset will then be used to train and compare machine learning models (artificial neural networks, random forest regression, support vector regression, gradient boosting and extreme gradient boosting), statistical indicators (coefficient of determination R², root mean square error RMSE, mean absolute error MAE and mean absolute percentage error MAPE) will be used to evaluate model performance and explainable artificial intelligence techniques will be applied to find out the relative contribution of mixture parameters to the predicted concrete performance. The framework proposed here establishes a reliable relationship between concrete composition and performance without requiring extensive experimental trial-and-error procedures. Laboratory experimentation, machine learning prediction, feature importance analysis and data driven optimization are expected to provide practical methodology for developing high performance nano-modified fiber reinforced concrete and more efficient sustainable concrete mix design.