ABSTRACT
Crack width control is essential for the serviceability of reinforced concrete (RC) structures, yet current design models rely on simplified assumptions that fail to capture strain localization and material heterogeneity. This study investigates uncertainty in crack width predictions using a finite-element model based on smeared crack and fracture mechanics approaches. The model incorporates the crack band concept to ensure mesh-objective energy dissipation and addresses limitations of experimental measurements, such as single-face observations and the exclusion of microcracks. Validation against experimental results of RC elements subjected to bending and direct tension examines the influence of mesh size, bar modeling, and concrete heterogeneity represented by random fields. The verification demonstrates that 3D modeling with embedded bars and an imposed minimum crack band size significantly improves prediction accuracy, achieving model uncertainty close to unity for maximum crack width. The findings provide practical guidance for applying smeared crack models in the serviceability verification of RC structures.
