Experimental results of passive vibro-acoustic leak detection in SFR steam generator mock-up

Regarding to GEN 4 context, it is necessary to fulfil the high safety standards for sodium fast reactors (SFR), particularly against water-sodium reaction which may occur in the steam generator units (SGU) in case of leak. This reaction can cause severe damages in the component in a short time. Detecting such a leak by visual in-sodium inspection is impossible because of sodium opacity. Hydrogen detection is then used but the time response of this method can be high in certain operating conditions. Active and passive acoustic leak detection methods were studied before SUPERPHENIX plant shutdown in 1997 to detect a water-into-sodium leak with a short time response. In the context of the new R&D studies for SFR, an innovative passive vibro-acoustic method is developed in the framework of a Ph.D. thesis to match with GEN 4 safety requirements. The method consists in assuming that a small leak emits spherical acoustic waves in a broadband frequency domain, which propagate in the liquid sodium and excite the SGU cylindrical shell. These spatially coherent waves are supposed to be buried by a spatially incoherent background noise. The radial velocities of the shell is measured by an array of accelerometers positioned on the external envelop of the SGU and a beamforming treatment is applied to increase the signal-to-noise ratio (SNR) and to detect and localize the acoustic source. Previous numerical experiments were achieved and promising results were obtained. In this paper, experimental results of the proposed passive vibro-acoustic leak detection are presented. The experiment consists in a cylindrical water-filled steel pipe representing a model of SGU shell without tube bundle. A hydrophone emitting an acoustic signal is used to simulate an acoustic monopole. Spatially uncorrelated noise or water-flow induced shell vibrations are considered as the background noise. The beamforming method is applied to vibration signals measured by a linear array of accelerometers on the shell. Satisfying results are obtained regarding to detection and localization of the source smothered by the background noise.