System-level thermal-hydraulic models provide bulk quantities but cannot directly resolve rod-wise temperatures or hotspot migration during local accidents. This study develops an FMI-based multiscale workflow that couples a Modelica primary-loop FMU with an OpenFOAMderived reduced-order model of a 3×3 lead-bismuth eutectic rod bundle through a Python driver. Four scenarios are evaluated: baseline operation, single-rod local overpower, and local overpower combined with mild or moderate loss of flow. The coupled model predicts bundleaverage outlet temperature, nine rod temperatures, hotspot temperature, and limiting-rod location. The hotspot-tooutlet temperature difference increases from 0.803 K in the baseline case to 18.274 K under local overpower with moderate flow reduction, while the hotspot migrates to the south-east rod. OpenFOAM calculations at 15 s and 40 s identify the same limiting rod. All four 60 s scenarios require 0.319 s of online execution time, demonstrating faster-than-real-time local accident assessment.