Fuzzy-based Acoustic Leak Localization at Complicated Structures

A method for detecting and localizing leaks at complicated three-dimensional topolo-gies by measuring the leak induced structure-borne sound and by applying fuzzy pattern classification procedures has been developed. The sound patterns necessary to train the fuzzy pattern classifier are generated with simulated leaks at the original structure. RMS-values of acoustic emission sensors are used as features for char-acterizing the occurrence and the location of a leak. Fuzzy pattern classification is applied for check and calibration of the sensor lines and for leak localization as well. Several types of neural networks and the fuzzy pattern classification are compared concerning the capabilities of classification.