Field of the invention
[0001] The present invention relates to a method of automatic detection of a synchronous
rotor-stator contact in operation of a turbine and relates to the field of measurement
and evaluation of mechanical oscillations and resonances of turbine rotors. It may
be applied for use in steam and gas turbines.
Background of the invention
[0002] During operation of the steam and gas turbines, an undesired contact between a rotor
and a stator may occur in some situations (called "rotor-stator rub" or "rubbing").
Said situations may occur e.g., during run-up of the turbine when vibrations are usually
highest while overcoming rotor natural frequencies. In the first phase of this contact,
seals fitted between the turbine rotor and stator are brushed first due to which the
amount of leaking medium increases and the turbine performance reduces. In the case
the rotor-stator contact is not early detected and removed through a suitable intervention
in the turbine operation, both the rotor and stator may get seriously damaged, or
rub may lead to the turbine failure with consequent huge financial loss. Early contact
removal may be achieved, for example, by change to the speed during the turbine run-up
or run-down, or by change to lifting oil parameters in use of the turbine on a swivelling
equipment.
[0003] The rotor-stator contact may be a partial, which refers to very short but at least
several times repeated contact between the rotor and the stator, or a full-annular,
which refers to continuous or almost continuous contact between the rotor and the
stator. In doing so, the full-annular rotor-stator contact is always preceded, at
least for a short moment, by the partial rotor-stator contact. At present, detection
of the partial rotor-stator contact is based in particular on offline analysis of
vibration signals, when the turbine operator monitors total level of the measured
vibration, as well as phasor of a first harmonic component in the vibration signal.
A phasor refers to visualisation of an amplitude of a harmonic signal and an initial
phase with respect to a key-phasor (initiation of the rotor revolution) in a complex
plane. In the case of a jump change in total vibrations or the phasor rotation of
the first harmonic component with a time varying or periodically changing amplitude,
the rotor-stator contact is detected, and a detailed data analysis is carried out
after finishing the measurement to eliminate a possibility of false-positive detection.
Therefore, this approach is inappropriate for the rotor-stator contact detection during
real use of the turbine, and may be essentially employed only for laboratory or experimental
purposes.
[0004] Another approach for detection of the partial rotor-stator contact is based on that,
in addition to change to the phasor of the first harmonic component of the vibration
signals, the contact is also accompanied by forming of sub-harmonic components of
the spectrum, with frequency corresponding to frequency of the rotor-stator impacts.
However, a disadvantage of the rotor-stator contact detection based on monitoring
of such sub-harmonic components is that frequency of some of them are very close to
the frequencies corresponding to manifestations of other defects, e.g., instability
of oil film of a journal bearing. Consequently, there may be a false positive detection
of the rotor-stator contact, or on the contrary, an erroneous evaluation of the manifestations
of the contact as a defect of a different kind. Solution of the above problem is disclosed
in
EP 2,746,541, which proposes a method suitable for detection and localization of the partial rotor-stator
contact in use of the turbine.
[0005] In addition, the patent and non-patent literature discloses other methods for detection
and/or localization of the partial rotor-stator contact based on quite different principles.
For example,
EP 1,533,479 discloses a method for detection of the rotor-stator contact where operational parameters
of the turbine, such as for example shell temperature, rotor eccentricity, condenser
pressure, and generator load are monitored. In doing so, the applicable algorithm
evaluates independently the change to individual values and in the case of an abnormal
or a jump change of any of them, the change is evaluated as a symptom of the rotor-stator
contact. A disadvantage of the method is that a jump change to a monitored value may
relate to other phenomena or faults that normally occur in use of the turbine. Hence,
there is very high likelihood of a false-positive rotor-stator contact detection.
Considering the dynamics of the monitored values and foreseen sensitivity of the algorithm
in use, the evaluation of change to these values followed by the rotor-stator contact
detection may not be contemplated as a sufficiently fast method applicable in real
in use of the turbine.
[0006] US 2008/240902 discloses a method for the rotor-stator contact detection that consists in monitoring
of turbine stator temperature on at least of a part of circumference thereof. A local
temperature increase, if any, is attributed to rubbing between the rotor and the stator,
and evaluated as presence of the rotor-stator contact. The point of contact on the
stator may be roughly localized based on placement of a temperature sensor that senses
the temperature increase. A disadvantage of the procedure is that at the moment when
sufficiently high heat amount is produced through the rubbing between the rotor and
stator for the temperature sensor to detect it, up to order of minutes may lapse since
rotor-stator contact origin. Over that time, the rotor, stator, or other turbine components
may get heavily damaged, and the rotor bending is unpreventable in a manner other
than the machine shut-down. Another disadvantage is that in case of more severe rotor-stator
contact, which occurs along the full circumference of the stator or a larger part
thereof, the stator temperature increase is in fact recorded evenly by most of or
all temperature sensors placed in the stator. Owing to this, the rotor-stator contact
is not always determined. A substantial disadvantage is that additional installation
of a higher number of temperature sensors into the turbine stator body requires shutting
down the turbine and high costs.
[0007] In addition,
US 2009/0003991 discloses a method for the rotor-stator contact detection that consists in continuous
measurement of blade clearance from stator and/or rotor blades height. A disadvantage
of the method, similarly to previous case, is the necessity of additional installation
of special sensors inside the turbine stator and related higher total costs. The contact
between a rotor blade and stator may be detected by measuring the change of gap between
the rotor blade and seal in the stator. Insofar as the clearances between the blade
and stator are higher than clearance between a shaft and the stator, the rotor-stator
contact primarily occurs on the turbine shaft. Real likelihood of successful and early
detection of the rotor-stator contact in this way is very limited. Another disadvantage
is possibility of false-positive detection of the rotor-stator contact in case of
change to dimension sensed by the sensors, which may occur also for different reasons,
e.g., due to rotor eccentricity, change to rotation speed, and more. Document
US 2021/140851 discloses a system and a method for automatic diagnosis of a power generation facility.
The system includes a data measuring unit for acquiring vibration data from a rotating
body of a power facility, a signal processing unit for signal-processing acquired
vibration data, and extracting and quantifying predetermined characteristic factors
with respect to a time domain, a frequency domain, and a shape area. It further includes
a characteristic pattern storage unit for storing a characteristic factor pattern
classified for each failure type, and a failure diagnosis unit for diagnosing whether
a power facility to be diagnosed has a failure. The failure type of the power facility,
on the basis of a classified characteristic factor pattern, is determined.
[0008] Document
US 4,478,082 discloses a method and an apparatus for detecting the occurrence of rubbing in a
rotary machine. In this method, at least one acoustic sensor is mounted on at least
the rotor part or the stator part of the rotary machine. The output signal from the
acoustic sensor is detected and then filtered to extract a frequency component having
a frequency substantially equal to the rotational frequency of the rotary machine.
The purpose of these steps is to detect the occurrence of rubbing in the rotary machine
as early as possible and also to locate the source of the rubbing. Even when the occurrence
of abnormal metal-to-metal contact at a bearing of the rotary machine is detected,
it is discriminated from rubbing so that the occurrence of the rubbing can be reliably
detected.
[0009] The objective of the present invention is to propose a new method for detection of
synchronous rotor-stator contact in use of a turbine, which eliminates the disadvantages
of the prior art, and allows for the rotor-stator contact detection in real time as
earliest as it occurs. In addition, the objective is to eliminate installation of
new sensors inside the turbine body, or inside flow part of the machine, and use the
existing (normally installed) sensors.
Summary of the invention
[0010] The present invention discloses a method of automatic detection of a synchronous
rubbing in a turbine. It is applicable both for gas as well as steam turbine. For
the method described herein, at least two sensors of rotor vibrations arranged on
the turbine stator, perpendicularly each other within one measurement plane are used.
The measurement plane is perpendicular to a rotor rotation axis. It means that the
sensors measure one area of the rotor, of which points are at the same distance from
ends of the rotor shaft. The described method for automatic detection of synchronous
rubbing may be carried out on multiple points of a single rotor at the same time.
It means that a plurality of measurement planes with relevant sensors of rotor vibrations
are employed, wherein measured data is processed and evaluated separately for each
measurement plane. A detection in a single of the employed measurement planes is enough
to detect the synchronous rubbing. A plurality of the measurement planes is favourably
employed for longer rotors where more accurate detection of the synchronous rubbing
is achieved.
[0011] The method is carried out that each sensor continuously measures the relative distance
of the rotating rotor and turbine stator in the place of the sensor. A trajectory
of eccentric rotor motion in the measurement plane is reconstructed by combination
of rotor distances filtered at rotational frequency (Ω) in an orthogonal coordinate
system. The rotational frequency is denoted 1X, and sometimes also referred to as
a first harmonic signal component. Furthermore, the trajectory of eccentric rotor
motion is divided into individual orbits. Each orbit includes data per one rotor revolution,
wherein the data includes information about the orbit origin coordinates in the orthogonal
coordinate system for the beginning of the rotor revolution. In the case of perfectly
centric rotor there should be no vibrations during rotor rotation, no eccentric motion
of the rotor in the measurement plane would be measured by rotor vibration sensors,
and no orbits would form. This is rather a theoretical case; in fact, vibrations and
eccentric rotor motion are always detected.
[0012] An initial amplitude defined by the distance of the orbit origin from the origin
of the orthogonal coordinate system is determined for each orbit. Then, an initial
angle is defined by angle between straight line defined by the origin of the orthogonal
coordinate system and the orbit origin and any axis having a known position in the
orthogonal coordinate system. It is favourable if an axis having known position is
the X axis or Y axis of the orthogonal coordinate system. The origin of the coordinate
system lies in axis of non-vibrating rotor, or lies in the centre of a circle inscribed
to the rotor journal bearing inside the stator.
[0013] Furthermore, the development of change to coordinates of the orbit origin over time
is monitored by monitoring the changes of the initial amplitude and the initial angle
between consecutive orbits. Of the measured values, the synchronous rubbing is detected
by cumulative meeting of these indicators for a unit of time:
- development of change to the initial angle of opposite direction of the rotor rotation.
It means that development of change to the initial angle is in the area of negative
values for the counter clockwise rotation of the rotor, and in the area of positive
values for clockwise rotation of the rotor;
- exceeding of the threshold value of the initial amplitude derivation;
- exceeding of the threshold value of the initial angle;
- exceeding of the threshold value of the area under curve, which is defined by the
coordinates of the origins of the consecutive orbits.
[0014] Meeting of only of some of these 4 points may indicate an impending synchronous rubbing,
however, the information is preliminary only. Evaluation of only some of these 4 points
as synchronous rubbing would result in high number of false-positive detections which
is not desired.
[0015] Generally, the primary excitation force acting on the rotating rotor is the centrifugal
force which acts in the point of a residual unbalance of the rotor. The centrifugal
excitation force is a harmonic function of time with frequency equal to the rotor
rotational frequency. The point where the centrifugal force acts is known as Heavy
Spot. Vibration response of the rotor to acting centrifugal force is the harmonic
function of time that is phase-delayed with respect to the excitation force. A place
on the rotor surface being steadily farthest from non-deflected (ideal) rotor centre,
or from the origin of the orthogonal coordinate system, is referred to as High Spot.
This place is angle-shifted counter the rotation direction when compared to the point
of the centrifugal force. The shift angle size between High Spot and Heavy Spot depends
on rotational speed of the rotor, and is defined by phase characteristics of a mechanical
system. In the case when rub occurs due to excessive vibrations, it usually occurs
in High Spot. In the case of the synchronous rubbing, which occurs periodically with
each rotor revolution, there is transient temperature rotor bending caused by rubbing
the stator and by local rotor temperature increase. The rotor bending may be interpreted
as another unbalance, and a point of another centrifugal force is formed in the High
Spot direction. Combining the centrifugal force acting in the point of original mechanical
unbalance and the centrifugal force acting in the point of new temperature unbalance,
a new effective centrifugal force is formed, which is, however, shifted compared to
original centrifugal force acting in the point of the mechanical unbalance by angle
α. The change of the initial phase of an excitation force acting on the rotor also
invokes a change to the rotor response initial phase 1X because the angle formed between
High Spot and Heavy Spot is defined for said rotor rotational speed by phase characteristic
of the mechanical system. The amplitude of the resultant of the centrifugal forces
will be higher or lower than the amplitude of original centrifugal force acting in
Heavy Spot considering the mutual angle between the original centrifugal force and
the centrifugal force with its point in High Spot. Change of the excitation force
amplitude invokes corresponding change to the 1X amplitude of rotor vibration response.
Should we assume that the synchronous rubbing occur in High Spot, we need to know
phase characteristics of the rotor response in each measurement plane to decide how
the rotor response changes in the rubbing in High Spot.
[0016] Let's assume that there is an acute angle between the centrifugal forces acting in
the points of Heavy Spot and High Spot for said rotor rotational speed. In the case
of occurrence of synchronous rubbing, the resultant of the original centrifugal forces
and the centrifugal forces that occurred due to temperature rotor bending will have
amplitude higher than amplitude of the original centrifugal force, and the initial
angle of the 1X orbit origin will change counter the rotor rotation direction. The
initial angle of the 1X orbit origin will change counter the rotation up to mutual
180° angle between the original centrifugal force and the centrifugal force in the
High Spot.
[0017] The method described in the present invention assumes measurement of relative rotor
vibrations in two mutually perpendicular directions. The angular position of the first
of the sensors in the rotation direction with respect to the horizontal axis may be
arbitrary ϕ
m (Fig. 1). Furthermore, a rotor key-phasor signal measurement is assumed. The rotor
key-phasor is one specific point of the rotor in the measurement plane. Rotor key-phasor
measurement is an assumption for dividing the measured signals into individual rotor
revolutions. It is usually carried out by an induction sensor located on the turbine
stator.
[0018] Then, the initial amplitude and the initial angle of the 1X orbit origin is evaluated
for signals of the relative rotor vibrations, e.g., by applying the order spectrum
calculation from said signals, or by estimation using the least squares method.
[0019] To detect the synchronous rubbing, the initial amplitude and the initial angle of
the 1X orbit are monitored (Fig. 2). The initial amplitude and the initial angle of
the 1X orbit origin are evaluated from the amplitudes and the initial 1X phases of
both signals of the relative rotor vibrations:

[0020] The method of automatic detection of synchronous rubbing is based on evaluation of
the following criteria (Fig. 3):
- exceeding of the determined threshold value of change to the initial amplitude of
the 1X orbit origin for a selected unit of time;
- exceeding of the determined threshold value of change to the initial angle of the
1X orbit origin for a selected unit of time;
- exceeding of the determined threshold value of the curve area circumscribed by the
origins of consecutive orbits for a selected unit of time;
- development of change to the initial angle of opposite direction of the rotor rotation
for a time period limited by T[0] and T[n].
[0021] The threshold values for each criterion are determined as 400% of standard deviation
of said quantity from a sufficiently long data line (at least 1 hour) in use without
rubbing present. The determined time unit may be different for each criterion. The
assumptions for use of the described method include turbine operation at constant
speed and knowledge of phase characteristic of the system.
[0022] Owing to the present invention, it is possible to eliminate installation of new sensors
inside the turbine body, or inside flow portion, and the existing (normally installed)
sensors may be used. This substantially cheapens the installation of the monitoring
system of the synchronous rubbing and increases availability of long-term monitoring
not only for newly installed turbines but also for machines already in use (no machine
shutting down is necessary when installing the monitoring device).
Description of drawings
[0023] The exemplary embodiment of the proposed technique is described with reference to
the drawings, on which
Fig. 1 - is a scheme of possible position of sensors of the rotor vibrations against
the rotor with indication of the sense of rotor rotation while measuring the relative
rotor vibrations;
Fig. 2 - filtered 1X orbit with the illustrated origin of the orbit;
Fig. 3 - characteristics for development of quantities for detection of the synchronous
rubbing;
Fig. 4 - course of the initial amplitude and of the initial angle of the 1X orbit
origin in the presence of the synchronous rubbing;
Fig. 5 - detection criteria of synchronous rubbing with indicated threshold values;
Fig. 6 - the detected presence of synchronous rubbing identified, marked in bold (exceeding
of all three threshold values and development of change to the initial angle counter
the rotor rotation direction).
Exemplary embodiment of the invention
[0024] Two sensors of the rotor vibrations are used in this exemplary method of the automatic
detection of synchronous rubbing in the steam turbine. The sensors are arranged on
the steam turbine stator perpendicularly each other within a single measurement plane
perpendicular to the steam turbine rotor axis of rotation. Each sensor continuously
measures the relative distance of the rotating rotor and the steam turbine stator
in the place of the sensor. In this example, the rotor rotates counter clockwise.
A trajectory of eccentric rotor motion in the measurement plane is reconstructed by
combination of rotor distances filtered at rotational frequency 1X in an orthogonal
coordinate system. Furthermore, the trajectory of rotor motion is divided into individual
orbits. Each orbit includes data per one rotor revolution, wherein the data includes
information about orbit origin coordinates in an orthogonal coordinate system.
[0025] An initial amplitude defined by distance of the orbit origin from the origin of the
orthogonal coordinate system is determined for each orbit. Furthermore, an initial
angle is defined for each orbit by angle between a straight line defined by the origin
of the orthogonal coordinate system and the orbit origin and any axis having a known
position in the orthogonal coordinate system. In this case, the X axis of the orthogonal
coordinate system is the axis.
[0026] Furthermore, the development of change of the orbit origin coordinates over time
is monitored so that the change of the initial amplitude value and change of the initial
angle between the consecutive orbits are monitored. The synchronous rubbing is detected
by the development of change to the initial angle counter the rotor rotation direction
(counter clockwise in this example - the area of positive values) and by exceeding
of threshold values of derivation of the initial amplitude, initial angle, and area
under curve being defined by the coordinates of the origins of the consecutive orbits.
[0027] In this example, the threshold values are:
- initial amplitude: 0.12 [µm/min]
- initial angle: -0.0168 [rad/min]
- area under curve: 0.1 [µm2] while assuming the orbit initial point curve in 1 minute.