2. Measurements and analysis of diagnostic signals.
Question 2.6.
What methods of analysis that are used in vibroacoustical
monitoring and diagnostics can be considered as most perspective ones? The
question was asked by the students of the Saint Petersburg State Marine
Technical University, chair "The reliability and technical diagnostics".
The question is answered by
Natalia A. Barkova:
According to our experience some methods of signal analysis
can occur to be perspective in monitoring, but not in diagnostics. And vice
versa. So in my answer I shall separate the monitoring from diagnostics.
1. In monitoring the most efficient are the methods of signal energetic
parameters analysis. If necessary the signal is preliminary divided on frequency
or spatial components.
The main frequency method of separation the signals by components is the
spectral analysis. In this method the energetic evaluation of each component is
given (its amplitude, power).
For spatial separation of signals the spatial separation of vibration (noise)
measurement points, or signal correlation measurements in two or more points are
used. The second approach is used more often in acoustical diagnostics of
extended uniform medium, for example, to detect the leakage in the pipe-lines.
It enables to define not only their energetic characteristics of the signal
components, but also their spatial coordinates, and also to extract the useful
signals from the interference, increasing the accuracy of the energetic
parameters evaluation.
2. In diagnostic problems, especially in detection of incipient defects, the
energetic parameters have the secondary value. The main value has the form of
the signal. Just by the form of certain components it is easier to detect the
defects. An example is the use of shock pulse method to detect certain types of
incipient defects in the bearing.
Unfortunately, any deviation of the signal form from the simplest one, i.e. the
sinusoidal one, expands its frequency band. In real machines many vibration
(noise) signals of complicated forms are summarized, but as their frequency
bands can overlap it is impossible to analyze the signal form by some certain
general method. In this case it is necessary to use some specialized methods of
extracting the signals of a certain form from the sum of many other types of
signals. An example is the cepstrum (double spectrum) signal analysis that
responses on appearance in the summarized signal the harmonic components,
modulated by the frequency or amplitude by periodical process.
This method showed its comparatively good efficiency in diagnostics of the
rotating machine's units with many teeth or blades as it has higher sensitivity
to certain defects of teeth (blades), impellers, gears, magnetic part of the
electric motors.
For already three decades we design the most general methods of analysis of
complicated form real signals that are necessary to be preliminary extracted
from a sum of several signals, part of which have complicated form. It occurred
that the most universal approach is the analysis not of the signal itself, but
the variations of its root-mean-square value or power. So the analysis of such
type of signal is reduced to the analysis of the form of its envelope. In this
case naturally all the problems of the preliminary extraction of analyzed signal
from the sum of many others do not disappear and have to be solved. The simplest
way to solve these problems is to measure only the high frequency vibration
(noise) components in certain frequency band and in the immediate vicinity of
the source.
When you transfer to the enveloped signal its form became significantly
simplified. It enables to use more general methods of analysis. Three most
general methods can be used in this case. The first one - you can compare the
form of the signal just with the form of the standard for different types of
defects. The second - the spectral signal analysis, if the signals are
periodical and have not very complicated form. The third - the statistical
analysis of the signals.
Nowadays the spectral methods of enveloped signal analysis are used most widely.
But we cannot exclude that the other two methods in the nearest future will also
find their application in solving the diagnostic problems.
