As part of the sustainable use of nuclear energy within a carbon-free energy mix in combination with renewable energies, fourth-generation fast neutron reactors are crucial for closing the fuel cycle and controlling uranium resources. Ensuring the safety of such a sodium-cooled reactor relies for a significant part on the early detection of gas voids in their circuits. In these opaque and metallic environments, optical imaging methods are ineffective, making it necessary to develop innovative techniques.
This PhD project is part of the development of Electrical Impedance Tomography (EIT) applied to liquid metals, a non-intrusive approach enabling the imaging of local conductivity distributions within a flow.
The work will focus on the study of electromagnetic phenomena in two-phase metal/gas systems, in particular the skin effect and eddy currents generated by oscillating fields.
Artificial-intelligence approaches, such as Physics-Informed Neural Networks (PINNs), will be explored to combine numerical learning with physical constraints and will be compared with purely numerical simulations.
The objective is to establish refined physical models adapted to metallic environments and to design inversion methods robust against measurement noise.
Experiments on Galinstan will be conducted to validate the models and demonstrate the feasibility of detecting gas inclusions in a liquid metal.
This research, carried out at IRESNE Institute of CEA Cadarache, will open new perspectives in electromagnetic imaging for opaque, highly conductive media.