User's Manual

67 I 72
7 Appendices
Correction explained in detail 7.1
Leakage noise is generally a wild mixture of tones, each of which can be associated with
individual frequencies (tone pitches, oscillations). The distribution of this frequency content is
referred to as a spectrum.
First, the time series signal is recorded. These time series values can undergo a mathe-
matical transformation, Fourier analysis.
A particularly efficient computational technique of Fourier analysis is the FFT (Fast Fourier
Transform). This is used to calculate the frequency content and, in particular, identify the
presence of dominant individual frequencies. Using this information, it is possible, for
example, to draw conclusions on disturbances (e.g., electrical machines).
Frequencies are specified in units of Hz (Hertz) as the number of oscillations per second. For
leakage noises, one usually finds a uniform mixture of frequencies, primarily located in the
range from 50 to 1000 Hz.
The knowledge of the spectrum, i.e., of the frequency response, is particularly helpful for the
computational use of filters, by means of which one can separate the useful and the
interfering portions of the noise measurement spectra.
Correlation is a mathematical method for comparing two time series signals with one
another.
A leakage noise is simultaneously recorded by two sensors at different locations. The sound
emitted by the leak spreads in the water pipe at a defined sound velocity. If the acoustic
event were to be brief and occur only once, e.g., a blow with hammer, it would be simple to
measure the arrival at both receivers. In this case, one would only need to compare the
arrival times at the two receivers with one another. If these are the same, the sound has
traveled the same distance to the receivers, i.e., the sound source lies in the middle of the
pipe between the receivers.
If the arrival times are different, it is possible to calculate the distance to the two receivers
from the measured time difference. The sound arrives at the closer sensor before the more
distant sensor.
Leakage noise, however, is not a one-time acoustic event, but it is also not completely
monotonous. Small, irregular changes in sound intensity and frequency occur. These
changing signatures travel to the two receivers in the same way as a hammer blow and may
arrive at the sensors at different times.
Because the identification and comparison of an individual signature is not as simple as in
the case of a hammer blow, the two time series signals are first recorded for a period of time;
the time series includes many individually measured signal values.
The two time series signals are mathematically compared with one another, i.e., correlated
with one another. If the leakage sound arrived at the sensors simultaneously, the same
signatures are compared and the signals are then similar and the correlation is high.