Detection of simulated steam leak into sodium in steam generator of PFBR by argon injection using signal analysis techniques

Abstract

The Prototype Fast Breeder Reactor (PFBR) is pool-type, sodium-cooled reactor. PFBR has eight steam generators (SGs) in its secondary circuit. The availability of SG units in the secondary circuit is of critical importance, as this determines the entire plant availability. Constant exposure to high temperature and high pressure may result in water or steam leaks in the tubes of SGs by which sodium and water reaction will take place. The growth in a leak may lead to tube rupture and large pressure increase in the secondary circuit. So, an important requirement from the viewpoints of safety and economics is to detect leaks at the incipient stage. Hydrogen detection methods, which are currently used to detect the initiation of steam leaks to sodium, involve transport delays, as the hydrogen evolved during the leak has to reach the sensor location. It is possible to detect a leak of 1 g/s within 1 s by acoustic leak detection. Another advantage is that it is possible to locate the position of a leak by installing several acoustic sensors on a SG. Experiments to develop a suitable signal-processing technique were carried out in the Steam Generator Test Facility (SGTF); argon was injected into sodium at argon pressure of 2 to 10 MPa through a 0.5-mm orifice. Signal-processing techniques, an autoregressive noise variance technique, and a wavelet detailed coefficient variance technique were studied and compared for their sensitivity to detect a steam/water leak. The acoustic technique employing wavelet decomposition is found to be promising for detecting a leak at the incipient stage. This paper discusses the details of the experiments carried out and the instrumentation and signal analysis techniques used for leak detection in SGs.