PROCEDE DE DETECTION DE CAVITATION

METHOD FOR DETECTING CAVITATION

Abrégé français

L'invention concerne un procédé de détection de cavitation pendant le fonctionnement d'une machine hydraulique, qui comprend au moins une roue à aubes, à l'aide d'au moins un capteur de bruit de structure. Le procédé comprend les étapes suivantes consistant : à détecter le signal de l'au moins un capteur pendant une durée qui comprend au moins une révolution de la roue à aubes; à filtrer le signal au moyen d'un filtre passe-bande; à effectuer un traitement sur le signal; à diviser temporellement le signal en intervalles de temps appropriés (les trois étapes successives sont effectuées séparément sur les signaux partiels individuels); à effectuer une décomposition spectrale; à appliquer un compteur d'événements dans le domaine de fréquence; à pondérer les événements relativement à leur étendue dans la bande de fréquence; à calculer au moins une valeur caractéristique à partir des événements pondérés de tous les signaux partiels; et à comparer l'au moins une valeur caractéristique avec au moins une valeur seuil.

Abrégé anglais

Disclosed is a method for detecting cavitation during operation of a hydraulic machine which has at least one impeller, said method using at least one structure-borne noise sensor, wherein the following method steps are carried out: detecting the signal of the at least one sensor over a period of time which comprises at least one rotation of the impeller; band pass filtration of the signal; preparing the signal; temporally dividing the signal into suitable time intervals (the following three steps are carried out separately for the individual partial signals); spectral decomposition; applying an event counter in the frequency domain; weighting the events with respect to their extent in the frequency band; calculating at least one characteristic value from the weighted events of all the partial signals; comparing the at least one characteristic value with at least one limit value.

Revendications

9
Patent claims
1. Method for detecting cavitation during the
operation of a hydraulic machine, which has at
least one impeller, with the aid of at least one
structure-borne sound sensor, characterized in that the
following method steps are carried out:
detection of a signal of the at least one sensor over
a time period which comprises at least one rotation of
the impeller;
bandpass filtering of the signal;
preparation of the signal;
temporal division of the signal into suitable time
intervals, wherein the next three steps following this
are carried out separately on the individual partial
signals: (1) spectral decomposition; (2) application of
an event counter in the frequency domain, wherein an
event is present if a signal point lies above a fixed
threshold value, wherein a number of successive signal
points above the threshold value count as a single event;
(3) determination of the extent of the events in the
frequency band; calculation of at least one
characteristic value from the number and the extent of
the events of all the partial signals, wherein a metric
which contains a multiplication of the number of events
with the extent of the events is used; comparison of the
at least one characteristic value with at least one limit
value.

Description

CA 02974847 2017-07-25
WO 2016/119939
PCT/EP2015/076500
- 1 -
Method for detecting cavitation
The present invention relates to the operation of a
hydraulic machine with an impeller, in particular a
turbine, pump or pump-turbine in a hydroelectric power
plant.
Cavitation can occur during the operation of a
hydraulic machine. Cavitation generally occurs in
operating ranges that are outside the optimum operating
range for which the hydraulic machine was primarily
designed. Cavitation may lead to increased wear of the
hydraulic machine. In certain operating states, the
occurrence of cavitation cannot be reliably predicted,
especially in the transitional areas of no cavitation -
cavitation. Many circumstances influence cavitation
behavior. Among the factors responsible for
discrepancies in the prediction are the influence of
air pressure, water temperature, sediment concentration
in the water or the degree to which the hydraulic
surfaces are worn down. At the same time, on the one
hand the exact dependence of the cavitation on the
parameters mentioned is often not sufficiently known,
and on the other hand some of the parameters cannot be
detected with sufficient accuracy. Therefore, in order
to exclude cavitation reliably, certain operating
ranges must be avoided. As a result, the usable
operating range of a hydraulic machine is restricted.
Another strategy for reducing the harmful effects of
cavitation is pursued in the case of some turbines, in
that air is fed into the water at suitable locations.
The feeding-in of the air generally requires
compressors, which have a not inconsiderable energy
consumption. Since, as described above, it cannot be
accurately predicted in which operating ranges harmful
cavitation will occur, in the case of these turbines
the blowing-in of air is already activated as a

CA 02974847 2017-07-25
WO 2016/119939
PCT/EP2015/076500
- 2 -
precaution in operating ranges in which still no
cavitation or little cavitation actually occurs. This
gives rise to an energy consumption that is greater
than would be required on the basis of the cavitation
behavior of the hydraulic machine.
The adverse effects mentioned could be avoided if
incipient cavitation could be reliably measured.
The object of the present invention is therefore to
provide a method for measuring cavitation that is
reliable and can be easily applied and can be used
during the operation of the hydraulic machine in order
to detect incipient cavitation and to choose the
operating state in such a way that cavitation is just
avoided.
Various methods for detecting cavitation by means of
measuring instruments are known. One possibility is for
example the use of structure-borne sound sensors that
are designed specifically for very high frequencies
(100 kHz to 1 MHz) and are attached to the turbine
casing. For further processing, it has proven to be
beneficial to form two characteristic values (W. Knapp,
C. Schneider, R. Schilling "Ein Monitor-System zur
akustischen KavitationsUberwachung von Wasserturbinen"
[A monitoring system for the acoustic cavitation
monitoring of water turbines], 8th international
seminar on "Wasserkraftanlagen" [hydroelectric power
plants], TU Wien [Vienna University of Technology],
1994). Characteristic value 1 represents the sum
effective value of the highpass-filtered time signal.
Characteristic value 2 is a counting signal. The
evaluator in this case counts the peaks in a predefined
time window.
The inventor has thoroughly investigated variants of
the measuring and evaluation methods that are known and

CA 02974847 2017-07-25
WO 2016/119939
PCT/EP2015/076500
- 3 -
described in the previous paragraph. In its
investigations, it found that cavitation can be
detected with these methods. However, limit values for
this must be determined for each hydraulic machine in
which cavitation is to be detected. If the
characteristic values calculated from the measured
values exceed the limit values, there is cavitation.
Investigations that were carried out on 4 different
hydraulic machines gave different limit values for each
of these hydraulic machines. The limit values on these
machines could only be determined because it was known
for these machines, or was otherwise possible to
observe, when cavitation occurs and when it does not.
During the operation of a randomly chosen hydraulic
machine, it is neither known nor can be otherwise
observed when cavitation occurs. Therefore, the
measuring and evaluation methods that are known and
described above cannot be applied in practice to a
randomly chosen hydraulic machine, since the required
limit values for the hydraulic machine are not known
and cannot be determined without an extremely great
effort (for example by fitting windows in the water-
carrying channels through which the occurrence of
cavitation could be visually observed).
The inventor has recognized that the stated object can
be achieved by finding an alternative measuring and
evaluation method which indicates cavitation by the
measured variable exceeding a limit value, it being
possible for the limit value to be chosen such that it
is the same globally for any randomly chosen hydraulic
machine. This global limit value can then be fixed on
one or more hydraulic machines that are well known
and/or accessible for viewing, and be used for the
measuring method on a randomly chosen hydraulic
machine.

CA 02974847 2017-07-25
WO 2016/119939
PCT/EP2015/076500
- 4 -
The stated object is achieved by the features of claim
1.
The way in which this is achieved is based on a novel
and inventive evaluation method, which is described
below. The measuring and evaluation method according to
the invention is likewise based on measurement with the
aid of structure-borne sound sensors. The basis for the
evaluation method that is used is the raw signal from
one or more structure-borne sound sensors. The signal
of the at least one sensor is detected over a time
period which is of such a length that it comprises at
least one rotation of the impeller. Proceeding from
this raw signal, the following evaluation steps are
carried out:
Choice of the frequency band to be considered: the
detected structure-borne signal is filtered with a
bandpass filter. Filtering limits that have proven to
be practicable are f_min=100 kHz and f_max=300 kHz. The
method also still works however with filtering limits
that deviate from the values mentioned, for example
with f min=200 kHz and f max-600 kHz.
Preparation of the signal: if appropriate, the filtered
signal is scaled, in order to take into account the
gain factor of the sensor used. Subsequently, the mean
value of the signal is subtracted, in order to
compensate for a possible offset of the signal. This is
followed by denoising the signal with a suitable
filtering function.
Division of the signal with respect to the position of
the impeller of the hydraulic machine: during the
measurement, the impeller of the hydraulic machine
rotates. The known rotational speed can be used to
calculate the time interval Delta T 360, in which the
impeller has turned once through 3600. The time

CA 02974847 2017-07-25
\
=
WO 2016/119939
PCT/EP2015/076500
- 5 -
interval Delta _ T _360 is then divided still further and
the denoised signal is considered separately in these
time intervals. The inventor has found that it is
expedient to carry out the division of the signal into
time intervals that correspond to an impeller rotation
of 10, i.e. that the corresponding time interval is
Delta _ T _ 1 = Delta _ T _360 / 360. The method also works
however if time intervals that are greater or smaller
are taken, for example for an impeller rotation of 0.1
to 100. With smaller time intervals, the computing
effort required increases and, with greater time
intervals, there is a deterioration in the resolution
with respect to the impeller position and in the
sensitivity of the measuring method because measuring
is performed integrally over a greater rotational angle
of the impeller. The subsequent evaluations up until
the calculation of the characteristic value are carried
out on the partial signals divided with respect to
these time intervals.
Spectral decomposition of the prepared partial signals:
the signal is spectrally decomposed for each time
interval, and the spectrum obtained provides
information on the spectral composition of the signal.
Application of an event counter: an event counter is
applied to the respective spectrum. For this purpose, a
threshold value must be fixed. This threshold value is
global, i.e. not dependent on the hydraulic machine
being measured. Details of fixing the threshold value
are described further below. There is an event to be
counted if at least one evaluation point comes to lie
above the threshold value thus fixed. If a number of
successive points lie above the threshold, there is
likewise a single event, which however is given a
higher weighting, as described below.

CA 02974847 2017-07-25
A
WO 2016/119939
PCT/EP2015/076500
- 6 -
Calculation of the characteristic value: the
calculation of the characteristic value is based on a
weighting of the individual events with respect to
their extent in the frequency band. The extent of the
events in the frequency band is determined by the
directly successive points in the frequency band that
lie above the threshold value being counted. The
falling of the amplitude below the threshold value
consequently indicates the end of the event. Many
metrics are conceivable for the weighting of the events
with respect to their extent. The inventor has used the
following metric: the counting values Zl and Z2 are
obtained for each spectrum: Z1 is the number of events
and Z2 is the number of points in the frequency band
that lie above the threshold value. The sum of Z1 and
Z2 is formed over all the spectra considered. The two
sums are multiplied. The characteristic value is the
product of these sums, but normalized to one rotation
of the hydraulic machine. As stated, other suitable
metrics may also be used.
Comparison with the limit value: the characteristic
value is compared with a limit value. The limit value
is the same for all hydraulic machines. If the
characteristic value exceeds the limit value, there is
cavitation, and suitable means for avoiding damage can
be taken (for example altering the operating state or
blowing in air).
Global determination of the threshold value and the
limit value: the threshold value and the limit value
were fixed by the inventor on the basis of the 4
hydraulic machines investigated. The same threshold
value and the same limit value were obtained for all 4
machines. The absolute value of the threshold value and
the limit value only depends here on the gain factor or
the sensitivity of the sensors used and the parameters
of the evaluation method (for example on the frequency

CA 02974847 2017-07-25
\
WO 2016/119939
PCT/EP2015/076500
- 7 -
band considered and the temporal scanning rate of the
sensors). It is also directly evident from what has
been said above that the absolute values of the
threshold value and the limit value are dependent on
one another. Thus, similar results are obtained if for
example the threshold value is chosen to be somewhat
lower and conversely the limit value is chosen to be
correspondingly higher. Furthermore,
moderate
alterations in the choice of the limit value and the
threshold value have only a slight effect on the
information derived with respect to the state of
cavitation of the machine, since the characteristic
value increases extremely steeply at the beginning of
cavitation. Therefore, the invention described here is
not based on the absolute values of the threshold value
and the limit value but on the evaluation steps
described above. It is only possible by the sequence of
evaluation steps according to the invention that have
been described to fix globally valid pairs of values
for the threshold value and the limit value. Therefore,
only one possible pair of values for the threshold
value and the limit value shall be given here: for the
sensors used by the inventor, a possible threshold
value of 0.01 V2 (for the frequency band of 100 to 300
kHz) and a possible limit value of 100 were obtained.
It is also possible to work with a number of threshold
values and limit values. If for example it is wished to
indicate cavitation that is just beginning, a second,
lower threshold value and a second corresponding limit
value are chosen. For the occurrence of cavitation that
is beginning, the inventor fixed a threshold value of
0.01 V2 and a limit value of 10 (likewise under the
conditions mentioned in the paragraph above).
Finally, it should also be mentioned that the sensor or
sensors used must of course be attached at suitable
locations of the hydraulic machine (for example on the

CA 02974847 2017-07-25
WO 2016/119939 PCT/EP2015/076500
- 8 -
turbine cover or on the suction pipe wall in the direct
vicinity of the impeller).