P SL CLTBW D EA X 3 b07 | 2008-09-14

P SL CLTBW D EA X 3 b07 | 2008-09-14
Application Note
Complementary differential
expansion measurements with the
SKF Multilog On-line System DMx
By Marcel de Boer • SKF Reliability Systems
As described in Differential Expansion application note number
CM3073 EN, the SKF Multilog On-line System DMx (a distributed
vibration based machinery protection and condition monitoring
system) is specifically designed to perform critical measurements
used in the control of large steam turbine generator trains. Perhaps
the most important measurement is differential expansion.
Differential expansion monitoring measures the change in axial
clearances between the machine rotor and stationary casing caused
by thermal changes inherent in most machines. The primary
purpose of such monitoring is to prevent axial rub between rotating
and stationary parts, the consequences of which can be
There are many configurations for measuring differential
expansion. This application note discusses the common sensor
configuration of Complementary Differential Expansion (abbreviated
to “CDE” or “Comp Diff Exp”) which is considered together with the
appropriate SKF Multilog DMx channels setup and configuration.
Reasons for measuring complementary differential expansion
instead of straight differential expansion include:
The SKF Multilog On-line System DMx.
• The desired working range of one non-contact eddy current
probe is not enough. Using the CDE method, the measuring range
• Due to machine design, there is not enough radial clearance
available to mount a probe of sufficient diameter versus range for
a straight measurement.
How does complementary differential
expansion work?
In Figure 1, the complementary differential expansion
measurement uses two sensors viewing a collar on the rotor
assembly. The amount of differential expansion capable of being
measured is based upon the distance between both transducers, but
it can never exceed twice the working range of one transducer.
As the rotor thermally expands or contracts, the rotor target area
moves from the working area of one transducer (“a” for transducer 1
in Figure 1) into the working area of the other transducer “b” and
eventually out of the measuring range of the “a” transducer.
The maximum range that can be measured is twice the working
range of one sensor. If the probes are positioned in such a way that
the second probe takes over at the end of the first’s working range,
the maximum range will be achieved.
Linear range
Machine Case
Transducer 1
Transducer 1
Transducer 2
Transducer 2
Figure 3
Figure 1
However, the transducers are usually mounted closer to each other,
generating an area where both transducers measure the target. In
Figure 2, transducer 2 is mounted closer to the collar, which
generates an area “c” where both transducers measure the target.
A further consideration in transducer selection is the available radial
clearance (“f” in Figure 4) for mounting a probe. In general, the
longer the measurement range of the probe, the larger the
diameter. The probe diameter cannot be larger than the available
collar target area divided by 2.5. The probe shall not be affected by
any metal other than that of the probe area.
Machine Case
>2½ x probe
tip diameter
a = linear range
Transducer 1
Transducer 2
Figure 4
Transducer types
Due to the versatility of the SKF Multilog DMx, different types of
eddy current probes (ECP) can be accepted for CDE measurements:
Figure 2
In Figure 3, the situation is shown where transducer 1 measures
the gap, and transducer 2 is in saturation. As the rotor grows with
respect to the case, the gap between transducer 1 and the collar
increases as the gap between transducer 2 and the collar decreases,
the increase in transducer 1’s gap continues until it exceeds its linear
range. At the same time, the collar exceeds the linear range of
transducer 1, entering the linear range of transducer 2.
This point is referred to as the “cross over point,” or the point at
which the system stops observing transducer 1’s gap, and turns
control over to transducer 2. This method allows measurements of
two times the linear range of a single sensor.
• Normal ECP signals, retrieved from externally powered drivers,
where the signal will be in the range between –2 and –18 V DC.
• Normal ECP signals, retrieved from drivers which are powered by
the SKF Multilog DMx, the signal will be in the range between
–20.3 and –2 V DC.
• Direct ECP signals, where the probe cable is connected directly to
the internal digital driver of the SKF Multilog DMx. This internal
driver linearizes the “true” probe curve using a digital algorithm,
which results in a much larger usable measuring range than an
analog driver: “equivalent” voltages from –2 up to –30 V DC can
be observed. The “equivalence” comes from an input sensitivity
(e.g. 4 V/mm). The SKF Multilog DMx converts by calculation the
true curve to the desired straight line of 4 V/mm. In this case the
scenario shown in Figure 3 will still apply, except the ranges are
larger than analog driver systems.
Probe calibration
Before adjusting probes that measure any differential expansion
type, a calibration curve of each probe should be generated with a
target that has similar magnetic characteristics as the final target.
The preferred method for setting the parameters of the CDE
configuration will be the positioning of the rotor exactly between the
minimum and maximum clearance. With symmetrical danger
values, this position is located at 0 mm (or 0 mils).
Next step is the adjustment of both probes; the probe gap should
be adjusted close to the end of its linear range, approximately –16 V
with normal eddy current probes. A control calculation should be
made for the most extreme positions – the probes should always
measure a value lower than their Probe OK High limit (with eddy
current probes this value is –1.6 V).
Once the probes have been mechanically adjusted, the offset in
the SKF Multilog DMx for each of the probes can be zeroed. This is
done with the SKF Multilog DMx Manager configuration software. If
this zero point has to be displayed as a different value (in case nonsymmetrical danger values have to be applied) an additional overall
axial offset can be programmed. This last value is normally the
average of both danger values and, once set up, the SKF Multilog
DMx will show this value if the rotor is in the central position.
gg During
machine commissioning, it is advisable to show both individual probe
measurements together with the CDE result and compare these values. If possible,
monitor especially the region where the cross-over point becomes active.
The K-position
In practice, the above procedure can rarely be used on large steam
turbines, because rotors aren’t easily moved and maximum or
“danger value” positions cannot be reached because the clearance
between blades also depends on operating temperatures also, and
internal temperature increases mean that both stators and rotors
expand together.
So what happens for real? The shaft must be in a known physical
position when the stationary and cold turbine is opened up to set the
probes. There should be a physical marker provided by the
manufacturer that indicates a known position of the rotor in this
state. This is known as the K-position or Zero-position. This position
will have a differential expansion value of, for example, +0.5 mm.
The two probes are then adjusted so that the measured voltages
match the calibration sheets of both probes, with an end result by
calculation of +0.5 mm.
Calculation of the complementary differential
For the CDE calculation, the SKF Multilog DMx follows some rules in
which both probes are checked for their signal validity.
The probe that is configured for the active direction (if the gap
increases, the result becomes more positive) is checked first. If this
reading from the probe is not inside Probe OK limits, the CDE value
will be based on the probe configured as inactive. If the active
configured probe is indeed between Probe OK limits then, a second
check is done on the calculated value from this probe. If the value is
smaller than the absolute value of the configured offset, the result
will be derived from the active probe, otherwise the result is based
on the inactive probe (unless this probe is outside its Probe OK
gg Due
to the versatility of the SKF Multilog DMx (accepting signals from external drivers
or the internal digital drivers), different types of Probe OK checks can be applied. With
direct ECP inputs, the check can be based on voltage, frequency or a combination of
both. To apply the above mentioned rules, voltage checking should be enabled for CDE
The CDE value
Based on the above rules, and the setup of offset and overall axial
offset, the following formula applies to calculate the CDE value (ΔX),
where the polarity of the sensitivity depends on the active/inactive
configuration of that probe.
The sensitivity will be multiplied with –1 if the probe is configured
as active.
∆X =
measured value
sensitivity (X – 1)
+ Offset + “OverallAxialOffset”
Special considerations should be taken with the offset setup,
especially when external probe drivers are used. If the offset value is
too high, dangerous conditions occur because the behavior of
traditional analog drivers show a “flattening” of the signal near the
end of their measuring range. If the offset has been calculated
(based on sensitivity), the possibility occurs that the offset will never
be reached. If this occurs, the CDE value will be derived from the
active configured probe and stabilizes, even if the rotor continues
moving in the active direction.
Example setup of the SKF Multilog DMx for
complementary differential expansion
To cover different types of installations in one example, the following
factors should be assumed:
• “non standard” sensitivity
• asymmetrical danger set-points
• External drivers – as more care should be taken with this kind of
setup. Where possible, direct ECP probes using the SKF Multilog
DMx internal digital drivers are recommended, as they have much
higher accuracy and range, and so are more tolerant of user error.
Transducer 1
Transducer 2
–2.2 mm
Rotor short
+5.0 mm
Rotor long
Figure 5
All dimensions are in metric units.
For English units: 25.4 mm = 1 000 mil
• Probe: CMSS 68
• Driver: CMSS 668-5
• Usable range: 145 mils = 3.68 mm, from 0.38 (“d” in Figure 5) to
4.06 mm
• Sensitivity: 100 mV/mil = 3 937 mV/mm
• Collar thickness (“c” in Figure 5): 100 mm
Alarm setpoints:
Figure 6
• Danger rotor short: –2.2 mm
• Alert rotor short: –1.9 mm
• Alert rotor long: +4.7 mm
• Danger rotor long: +5.0 mm
• Maximum detectable range: 2 x 3.68 = 7.36 mm
• Range to detect: 5.0 + 2.2 = 7.2 mm
Once the parameters are set up (for example, the danger level of
–2.2 mm will be at |0.48| mm), the other values can be calculated.
For any kind of axial measurement, it is necessary to use a probe
calibration graph.
From the probe calibration graph (Figure 6), the real sensitivity
can be determined:
The difference between the maximum detectable range and the
required range is the value that can be located between the
transducers or added to the usable range.
The first option introduces an area of 0.16 mm, where both
sensors measure the differential expansion (the result will only be
derived from the active configured transducer).
• (-16.49 + 2) * 1 000/(4.06 – 0.38) = –3 937 mV/mm.
• Danger low = –2.2 mm = |0.48| mm ‘
• The average of both danger values = (–2.2 + 5) = 3.6 mm.
• The zero offset for transducer 1 (active setup, thus offset should
be negative) = 3.60 + 0.48 = -4.08 mm.
• Above result in an overall axial offset of –2.2 + 3.6 = +1.4 mm.
–– Extra check could be the zero voltage, but this is not used in
further calculations. The zero voltage = –1 * –4.08 * –3 937 =
–16 063 mV (active setup, sensitivity = negative).
• If transducer 1 is at 4.08 mm, the calculated CDE result is +1.40 mm.
a = 0.38 + 2.20 + 100 + 5.00 + 0.38 = 107.96 mm
The second option introduces an extra 0.1 mm clearance between
probe tip and collar, besides the 0.38 mm offset of the linear range
(d ‘ 0.48 mm).
A similar calculation should be done for transducer 2:
a = 0.38 + 0.10 + 2.20 + 100 + 5.00 +0.06 + 0.38 = 108.12 mm
• There was 0.06 mm left for transducer 2 clearance ‘
• The zero offset for transducer 2 (inactive setup, offset should be
positive) = 3.60 + 0.38 + 0.06 = +4.04 mm.
• Or zero offset = 2x maximum range – offset (1) = 2 * 4.06 – 4.08
= +4.04 mm. The zero voltage = –1 * 4.04 * + 3 937 =
–15 905 mV.
Next step will be the physical adjustments to
both probes:
1 The easy solution: Start the configuration
software; program both measurements
with above values and mechanically
position the rotor exactly in the middle
between both danger values. Adjust
transducer 1 until the software displays a
measured (CDE) value for transducer 1 of
(–2.2 + 3.6) = 1.4 mm. Adjust transducer
2 until the software displays a measured
(CDE) value for transducer 2 of (5 – 3.6) =
1.4 mm. The benefit of this procedure is
that both probes are in their measuring
range and an easy check can be
performed: the result of the CDE
Figure 7
measurement should be equal for both
transducers and has to be the initial value
+ the overall axial offset = 1.4.
2 It becomes more complicated when the collar is in another known
position (for example +0.5 mm). Transducer 1 can be adjusted
according to the above procedure, and should be ‘gapped’ at
0.5 mm, or –16.06 + (1.40 – 0.50) * 3.937 = –12.52 V.
As a result, two calibration graphs are generated with the actual
values (Figure 7). These graphs can be combined into one CDE
result graph that visualizes the measurements behaviour see
Figure 8.
Adjust channel 1 until result =
–12.52 V or +0.50 mm
Similarly, the position of transducer 2 is determined:
• For this example, it is assumed that the probe calibration graph is
the same as the one for transducer 1.
Based on transducer 2, a value of –15.91 V is equal to 1.4 mm, this
transducer would theoretically display a value of –15.91 – 1.40 *
3.937 = –21.42 V at 0.00 mm or –15.91 – 0.90 * 3.937 = –19.45
Volts at +0.50 mm.
For transducer 2, this result (–19.45 V) should be the value to
adjust if the collar is in +0.5 mm position, except that this voltage is
above the Probe OK value (or out of the linear range), therefore
another solution is necessary to adjust this sensor. Some of the
possible solutions are:
• Move the collar towards the transducer and measure this distance
with a dial indicator.
• Manufacture a metal shim (same kind of material as the collar),
measure the thickness of the shim, and subtract that value from
the –19.45 V.
• Adjust transducer 2 so that the distance between both
transducers is between 107.96 mm and 108.12 mm.
Figure 8
SKF Multilog DMx Manager software setup
Figure 9 to 16 show how to setup the channel properties for the
previous example in SKF Multilog DMx Manager.
Figure 9
Configuration for channel 2 is similar as that for channel 1, except
for “target positive direction” and the “offset”.
The programmed sensitivity should be based on the calibration
curve and will be unique for each probe.
Overall axial offset =
calculated cross over
Offset =
Voltage/Sensitivity =
16063/3937 = 4.08
Figure 10
Figure 13
If channel 1 is active then
channel 2 is inactive
Figure 11
Figure 14
Figure 12
Figure 15
Configuration of the measurement setup for the second CDE
Figure 18
Figure 16
The SKF Multilog DMx processing setup
Figures 17 to 20 show how to setup the processing of collected data
in SKF Multilog DMx Manager.
Figure 19
Figure 17
SKF Multilog DMx live data display
Figure 20
Common pitfalls in the setup of
complementary differential expansion
Regardless of the monitor type used (SKF Multilog DMx, SKF
M800A, VM600 or other vendor product) there are a number of
common mistakes that can be made when setting-up a CDE
installation. The CDE measurement is arguably the most critical
measurement a TSI system will make, so it has to be right!
Pitfall 2 – incorrect sensitivity
Another pitfall is the programming of an incorrect sensitivity.
For example, programming the monitor with the standard
sensitivity for 4140 steel (3 937 mV/mm), whereas the real target
material produces a different sensitivity. Figure 22 illustrates this
Pitfall 1 – calibration graphs
A familiar error is the skipping of the creation of calibration graphs.
Often the cross-over point is selected close to the Probe OK limits,
in order to achieve the widest possible measuring range. Figure 21
exaggerates the behavior of an eddy current probe to give a better
visualization of why this is a mistake without true calibration curves:
Figure 22
Once probe 1 measures –16 V, the monitor displays again a value of
3.9 mm. The real value will be 3.5 mm.
As the gap increases towards the “true” crossover at –17.6 V, the
measurement needs to be taken over by the probe 2. Then two
conditions can apply:
Figure 21
From Figure 21, the conclusion could be made that this probe
measures up to 7 mm (at –17.6 V). The programmed sensitivity
(green line) is 3937 mV/mm. If the crossover point has been
selected at –16 V, the SKF Multilog DMx displays a value of 3.9 mm,
while the real expansion is already 4.9 mm. If the rotor grows
further, the second probe ‘takes over and the measurement returns
to the real value. In the above example, an accuracy problem occurs
between –13 and –16 V for the first probe. Fortunately, this behavior
is not close to the danger set points.
This problem can be witnessed as a “jump” in the CDE value
reported. A real value of 4.9 mm will be shown as 3.9 mm, the
second probe takes over and the next reported value for the CDE
measurement will be 5.1 mm. CDE jumps from 3.9 to 5.1 mm.
To improve above measurement, a sensitivity correction can be
applied to both sensors, based on the blue thin line, where the
crossover point is set at the “true” crossover point of –14 V.
• If probe 2 is not yet between Probe OK limits, the end result will
still be derived from probe 1 and remain incorrect until probe 1’s
OK limit is passed.
• If probe 2 is within Probe OK limits, the end result will cross over
to probe 2. In this example, the value will be reported as 4.3 mm
while the real gap is 3.8. When gap increases further to 3.9 mm,
the CDE result falls back to 3.5 mm.
Also, combinations of incorrect sensitivity and calibration errors can
occur, together with incorrect programming of other parameters,
especially offset values.
Pitfall 3 – incorrect interpretation of the
crossover point
Different monitor systems detect the crossover point in different
ways. Some use the Probe OK detection limit of each sensor, others
use the average point between the offsets.
As the internal digital drivers of the SKF Multilog DMx offer
significantly longer probe measurement ranges (compared to
traditional probe systems), the crossover point is determined by the
offset of the active probe only. Once this offset is reached, the
reading switches to the inactive probe.
A problem occurs if this methodology is not accounted for when a
normal external eddy current probe system is used.
This is illustrated in Figure 23 (overstated for clarity, as in the
previous examples).
In this scenario the guidelines for configuring the SKF Multilog
DMx have been followed correctly, and the sensitivities have been
determined from real probe curves. The rotor is positioned between
Danger limits and the two probes are adjusted to reach each of their
offset values, and hence show a 0 value in the SKF Multilog DMx
Manager display software.
Adjust channel 1
until result = 0.00 mm
Adjust channel 2
until result = 0.00 mm
However, because of a “flattening” of the driver response, an
incorrect interpretation of the probe curve can occur, see Figure 23.
Figure 23
Transducer 1 measures –16.06 V and the displayed CDE is
1.40 mm. The shaft starts moving towards the probe tip until it
reaches the real danger value (–2.2 mm) following the curve of the
black line. At –2.2 mm, the measured voltage is –7 and therefore the
displayed CDE result will be –0.9 mm, not even close to the real
danger limit of –2.2 mm!
The problem is avoided by ensuring a correct calibration of probe
and driver to the true target material of the collar. This is easily done
with the inbuilt drivers of the SKF Multilog DMx which eliminate this
potential error because of the better linearization of the probe
calibration curve.
The complementary differential expansion for both probes display
the calculated +1.40 mm and the conclusion will be made that the
setup is correct.
For additional information on SKF Reliability Systems products, contact:
SKF Reliability Systems
5271 Viewridge Court • San Diego, California 92123 USA
Telephone: +1 858-496-3400 • FAX: +1 858-496-3531
Web Site: www.skf.com/cm
® SKF and Multilog are registered trademarks of the SKF Group.
All other trademarks are the property of their respective owners.
© SKF Group 2009
The contents of this publication are the copyright of the publisher and may not be reproduced (even extracts) unless prior written
permission is granted. Every care has been taken to ensure the accuracy of the information contained in this publication but no liability
can be accepted for any loss or damage whether direct, indirect or consequential arising out of the use of the information contained
herein. SKF reserves the right to alter any part of this publication without prior notice.
Publication CM3118 EN • March 2009
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF