Three-dimensional crack reconstruction from Beam–Particle Model for CFD-based leakage assessment

Accurate crack characterization in reinforced concrete structures, such as the containment walls of 1300 MWe nuclear power plants,is essential for reliable air leakage estimation. The complexity of crack patterns—including their path, opening, and surface roughness—significantly influences predictions of air leakage. Traditional modeling approaches, such as applying Poiseuille’s law tosimplified geometries, depend on indirect parameters like the tortuosity coefficient, which is challenging to predict accurately andhas limited applicability, thereby increasing uncertainty. This study introduces a novel numerical post-processing tool based on theBeam-Particle Model (BPM) approach, aimed at enhancing the precision of crack predictions derived from finite element simulations. It employs graph theory to detect micro-crack paths and reconstruct a possible macro-crack geometry, thus improving therepresentation of crack characteristics. The generated data can be integrated into computational fluid dynamics (CFD) simulationsfor more precise airflow modeling or used to calibrate simplified methods, such as Poiseuille’s law, with numerically obtained crackdescriptors. The fracture geometries produced by the tool in the Brazilian splitting test are validated against optical microscopicobservations, demonstrating the tool’s ability to capture crack tortuosity. In contrast, the PSD-based assessment of the Hurst exponent on BPM-generated crack surfaces yields non-physical values, so surface roughness could not be quantified numerically due toBPM computational time constraints. In addition, the current reconstruction algorithm is limited to a single dominant crack pathand cannot handle crack branching or macro-crack turnover, although it can detect separated macro-cracks. These results establisha benchmark for future efforts to incorporate detailed crack geometries into CFD-based air leakage estimates, underscoring thetool’s potential to reduce uncertainty in such predictions.