Leckdetektion an komplizierten dreidimensionalen Topologien

The well-known methods of the acoustic leak localization using the attenuation differences or propagation time differences (e.g. cross correlation analysis) are restricted to the linear geometry of pipe lines. At complicated three-dimensional topologies they fail because there are several modes of the structure-borne sound that have different propagation velocities and because numerous connecting branches generate multiple sound reflexion. Therefore, a new method for detecting and localizing the leaks at complicated geometrical structures was developed and experimentally proved at an original facility. This approach is based on the detection of the leak induced airborne sound or structure-borne sound as well as on the application of pattern recognition procedures. The efficiency both of Fuzzy Logic and neural networks for the acoustic leak localization could be experimentally proved. Though the method has been prototypically developed for Russian pressurized water reactors of VVER-440 type, it can also be used by the self-learning capability of the classifiers at other pressurized facilities.