Self-Sensing Concrete Using Fibers and AI: A Review-cum-Experimental Framework for Intelligent Structural Health Monitoring

Abstract

Concrete is the most widely used construction material on earth, yet conventional reinforced-concrete structures remain largely “blind” to their own internal condition, relying on periodic visual inspection or costly, discretely mounted external sensors for structural health monitoring (SHM). Self-sensing (also called intrinsic or “smart”) concrete addresses this gap by dispersing electrically conductive fibers and fillers — carbon fiber, steel fiber, carbon nanotubes (CNTs), and graphene nanoplatelets (GNPs) — directly within the cementitious matrix, so that the material itself reports strain, stress, and internal damage through a measurable change in electrical resistivity, a phenomenon known as piezoresistivity. This paper presents a review-cum-conceptual-experimental study of fiber-based self-sensing concrete integrated with Artificial Intelligence (AI) for real-time, data-driven SHM. The percolation behaviour, piezoresistive mechanism, mix-design considerations and testing methodology for carbon-fiber- and steel-fiber-doped self-sensing concrete are described, and a framework is proposed in which electrical-resistance / fractional-change-in-resistance (FCR) signals are logged, pre-processed, and fed into machine-learning and deep-learning models — Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), Transformers, and Physics-Informed Neural Networks (PINN) — for automated strain estimation and crack/damage classification. Representative results reported in recent literature, together with illustrative laboratory-style trends compiled for this study, indicate gauge factors in the range of 20–225, percolation thresholds near 0.5–1.0 vol.%, and AI-based crack-classification accuracies exceeding 90–97%. The paper closes with the practical applications, limitations, and future scope of AI-integrated self-sensing concrete for next-generation smart infrastructure.