Coastal nuclear power plants (NPP) face serious seismic and geotechnical hazards, in which severe ground shaking can be followed by softening, liquefaction and tsunami effects. Conventional discrete accelerometers measure the response of massive reinforced concrete structures at only a few points. They cannot eliminate internal deformation fields or the changing ground-foundation structure (SSI) interaction that determines the safety margin. This article is a theoretical overview and a conceptual proposal. It considers the possibility of combining three existing technologies into a single monitoring approach: (i) an RC steel reinforcement frame, considered as an integrated mechanically coupled waveguide and voltage carrier; (ii) radiation-resistant Fe-Ga (galphenol) magnetostrictive transducers attached to reinforcement rods; and (iii) artificial intelligence in particular, physics-based neural networks (PINNS) for interpreting signals. Based on the literature on magnetostrictive sensing, controlled waves and coupling properties in fittings, SSI and scientific machine learning, the theoretical feasibility of the transition from high-frequency ultrasonic monitoring of the condition of structures to the low-frequency seismic range has been assessed. An analysis was carried out to identify the main obstacles to the implementation of this concept. The feasibility criteria and a step-by-step verification plan for future work are described.

nuclear power plant,seismic monitoring,magnetostrictive sensing,Galfenol,physics-informed neural networks