Covariance Matrix based statistical features have been used to detect leak,
identify noise source and characterise noise transmission off-line on data
from PFR Steam Generator Units. The methods show promise in their ability
to detect leak in actual Steam Generator Units from the noise data with ver
y low S/N ratios. Further analysis of data have yielded useful information
on the potential of acoustic leak detection system. (C) 2000 Elsevier Scien
ce Ltd. All rights reserved.