Companies in the energy sector are increasingly considering the use of liquid metals in their operations (e.g. for cooling). This presents new challenges related to the inspection and monitoring of these infrastructures. Indeed, these industries often operate in very harsh environments (e.g. ionizing radiation) and under very high pressure and temperature conditions. Thus, the ultrasonic inspection methods normally used must be adapted to these new conditions. Moreover, methods based on machine learning can be used to optimize the processing of the received signals. The objective of this project is to optimize the propagation of ultrasonic waves in liquid metals and at their interface in order to collect data that will allow, after automated processing, better decision-making.