FICFB: A comprehensive framework for reliable bolt loosening monitoring in nuclear power plants under multi-source uncertainty

In nuclear power plants (NPPs), bolted joints serve as the most widely used connection structures, performing critical purposes such as fixation and sealing. Bolt loosening can compromise the structural integrity and sealing performance of critical components, potentially leading to coolant leaks or even catastrophic failure. Thus, reliably monitoring the bolt connection status is essential. However, this task is severely hindered by multi-source uncertainties in NPPs. Specifically, harsh reactor environments that prevent conventional testing, along with diverse bolt applications, create complex environmental and structural uncertainties, which restrict the generalization of current methods. To address these challenges, a bolt loosening monitoring framework based on improved characteristic frequency bands (FICFB) is proposed. It integrates two enhanced strategies: an improved modal analysis for scenarios where conventional testing is infeasible, and a similarity-based approach utilizing improved singular value decomposition (SVD), genetic algorithms (GA), and indicator evaluation for cases without operational excitation. Verified by a proportional reactor pump test bench under complex uncertainties, the framework reliably identifies four bolt connection statuses across two scenarios, achieving detection accuracies of 94% and 95%, respectively. The results demonstrate its strong potential for reliable monitoring in diverse NPP applications.