Courtesy of Kop-Flex/Emerson Power Transmission Corp. JPC3.
pdf
A NEW DEVELOPMENT IN CONTINUOUS TORQUE MONITORING COUPLINGS
Joe Corcoran
Manager, High Performance Applications
Kop-Flex, Emerson Power Transmission
Baltimore, Maryland
ABSTRACT
The use of continuous torque monitoring couplings is
quickly becoming an integral part of many predictive
maintenance programs in the petrochemical and process
industries. More and more operating facilities are using
instrumented torque-measuring couplings in order to know how
their critical equipment is performing so that the intervals
between scheduled shutdowns can be chosen appropriately.
This paper will cover a brief history of torquemeters; ie., the
basic measuring types which are 1) strain gage types and 2)
torsional deflection types. Discussed will be the theories of
operation on how they work. Also discussed will be the
advances and limitation of these types. Included will be the
accuracies that can be obtained by each one and the features
and limitation of each. In addition, covered will be a new
advance in the mechanical torsional deflection, or phase shift
type, which improves the overall design so that mass elastic
properties of the coupling with the torque meter are more
compatible with high speed turbomachinery.
INTRODUCTION
Performance Testing and Shaft Power Measurement
Field performance testing and monitoring is essential in
turbomachinery to assess its current condition. For example the
objective of field testing gas turbine driven compressor sets is
typically to verify acceptance criteria such as heat rate, specific
fuel consumption, turbine shaft power, and compressor gas
power. Generally, performance testing which is done to
determine if the machinery meets the manufacturers guaranteed
design points should be performed in the OEMs test facility,
where the accuracy of the instrumentation as well as the control
of environmental factors are better.
Proceedings of ASME DETC 2000
Power Transmission and Gearing Conference
September 10-13, 2000 Baltimore, Maryland
DETC2000/PTG-14455
Steve DErcole
High Performance Engineering Product Manager
Kop-Flex, Emerson Power Transmission
Baltimore, Maryland
However, field performance testing has become most useful
in establishing a baseline from which the machinerys future
health can be measured. Direct measurement of the shaft power
between connected machinery enables one to isolate which
machine is responsible for an overall decrease in output.
Continuous on-line monitoring of the machinerys output power
provides operational trending data. Predictions in performance
degradation can be made so that corrective action schedules can
be established. A detailed history of the relative performance of
the machinery is needed to accurately assess its condition. The
accuracy of these field measurements are critical to the
reliability of any Predictive Maintenance Program.
Factory testing of the equipment will typically involve
smaller test uncertainties than field testing. During factory
testing of gas turbines for example, the shaft power is usually
measured directly by coupling it to a dynamometer. Load cells
measure the reaction forces on the casings to provide a direct
measurement of torque so that shaft output power is taken as the
product of measured torque multiplied by the operating speed.
In field testing however, unless a torque measuring coupling is
used, the gas turbine shaft power is not able to be measured
directly. It must be determined by performing a heat balance
with the calculated gas power of the driven compressor, or by
performing an energy balance on the gas turbine system. Using
a torque metering coupling can achieve almost a similar
accuracy as the factory test method. The heat balance and
energy balance methods are subject to significantly higher
measuring uncertainties (Kurz, Brun, Legrand, 1999).
Heat balance and energy balance methods are dependent on
measurements of pressures, temperatures, flows, gas
compositions, and mechanical losses. Each of these measured
parameters has its own instrumentation tolerance, which
contributes to the overall test uncertainty. The largest
instrumentation tolerance is due to gas composition (up to 5%),
with other measurement errors due to pressure (up to 2%), flow
1 Courtsey of Kop-Flex Emerson Power Transmision Co
(up to 2%), equation of state (up to 2.5%), and temperature (up
to 4F). If the shaft output power is known, the gas turbine
heat rate and efficiency can be determined. If a torquemeter is
used, the total uncertainty for the gas turbine power can be
reduced to about one percent to 1.5 percent. (Kurz, Brun,
Legrand, 1999). Without a torquemeter, the error can be as
much as 7%. (R,D. and J.D. van Millingen, 1991).
Similarly, the accurate performance measurement of a
centrifugal compressor is very dependent on the quality of the
field data. Again an important parameter is the shaft
horsepower, which can be calculated directly if a torquemeter is
available. Otherwise, a heat balance method is recommended,
such as that given in ASME PTC 10 (1997). For example,
Equations 1 and 2, (Wilcox, 1999) are the general formulas for
calculating input shaft horsepower into a compressor. To get the
shaft horsepower in this manner is fraught with the attendant
measurement tolerances (errors).
. .
(m1 m S 1 )(h2 h1 ) + QR
SHP = + HPMECH (1)
2545
SHP = HPGAS + HPMECH (2)
NOMENCLATURE
SHP = Shaft horsepower
m1 = Mass flow into compressor (lbm/s)
m2 = Mass flow out of compressor
h2 = Specific enthalpy out of comp. (Btu/lbm)
h1 = Specific enthalpy into comp.
HPGAS = Compressed gas hp
HPMECH = Mechanical hp losses dissipated through seals,
bearings, etc.
QR = Radiative Heat Transfer
TORQUEMETER COUPLINGS
Methods Of Torque Measurement
The growing popularity of instrumented couplings for
continuous on-line torque monitoring has led to the widely used
term torquemeter (or torsionmeter) couplings. There are addition to torque, the strain gages of the Wheatstone Bridge
several varieties of instrumented torquemeter couplings
currently available. Each are capable of providing torque
measurement through non-contacting means so there is no
longer a need for the extra bearing supports associated with a
torquemeter of years past.
These torquemeters physically measure the torque being
transmitted between the two machines of which they are
connected. Since they also measure the speed, the typical
preferred output of these torquemeters is power (torque
multiplied by speed).
All torquemeter coupling designs are faced with the task of
detecting a physical change in the coupling due to torsion while
it is rotating, and getting this information to a stationary output
device (generally in the control room). Over the years many
methods have been devised to measure the torsional effects
exhibited by the coupling. These methods range from
measuring changes in the acoustics of coupling mounted piano
wires to the application of magnetic circuits which sense
changes in permeability as the coupling winds-up. Most of
these methods have fallen short of the accuracy required for
meaningful use as performance monitoring instrumentation.
The challenge for accurate and reliable torque
measurement is that each system is faced with determining
those physical changes associated with torque alone while the
coupling is subjected to a combination of torque, bending, and
centrifugal loads. (See Figure 1).
Discriminating the effects between these multiple loads has
boiled down to two basic methods of detection:
1) Measurement of localized torsional strain, and
2) Measurement of overall torsional deflection.
Strain Gage Type Torquemeters
There are several variations of the strain gage torquemeter
system currently available. Each of them operates on the same
general principle of:
1) Getting electrical operating power from an outside
source to the coupling
2) Feeding that power through a four arm strain gage
bridge located on the rotating coupling
3) Transmitting the resulting signal from the coupling
back to a stationary receiver.
The strain gages are usually directly affixed to either the OD
or the ID of a thinned down area on the coupling spacer (center
spool piece). (For slower speed applications, some
manufacturers provide a clamp-on split collar which contains
the strain gages). As torque is applied, the localized twisting in
the area of the strain gages creates a signal by the unbalancing
of the strain gage bridge. Since the coupling spacer will be
exposed to axial, centrifugal, and misalignment loads in
(Figure 2) must be mounted precisely at 45 from the couplings
axis in order minimize the strains from these extraneous loads.
2 Courtsey of Kop-Flex Emerson Power Transmision Co
 Figure 2. Wheatstone Bridge Used In Typical Strain Gage
Torquemeters

The rotating electronics condition the signal received from

the stationary component and feeds it through the strain gage
circuitry. (The rotating strain gage circuitry is usually provided
by the manufacturer with a protective wrapping due to the
sensitivity of the circuitry to handling damage and possible
chemical contaminates.). The output of the rotating strain gage
circuitry is amplified and transmitted back to the stationary
component either by an FM (frequency modulated) signal, or by
a second rotary transformer - depending on the manufacturer.

Figure 1. Torquemeters Must Decipher Torque Induced

Effects From Other Combined Loads.

In the past, the method of transmitting power to and

receiving signals from the rotating instrumented torque-
measuring coupling involved the use of contacting slip ring
arrangements. This rendered them useful for only low-speed,
high-torque applications and presented problems related to wear
and foreign particulates. Today, most strain gage torquemeter
systems have overcome these problems by using non-
contacting, electro-magnetic induction techniques.
The basic strain gage type torquemeter (Figure 3) consists of a
stationary component and a rotating component. Both
components contain electronics. The stationary component Figure 3. Typical Strain Gage Torquemeter Coupling
(stator) provides power to the rotating component (rotor) via
electromagnetic induction between windings contained on each
component. The air gap between the stationary and rotating From the stationary component, the signal is typically sent
windings allows for relative axial, angular, and offset type back to the control room as an industry standard analogue
excursions of the coupling during operation. signal for connection to the users data recorders or
programmable logic controllers (PLCs).

3 Courtsey of Kop-Flex Emerson Power Transmision Co

Torsional Deflection (Phase Shift) Type Torquemeters
As with stain gage types, there are several variations of the
torsional deflection torquemeter system currently available.
Each of them operates on the same general principle of
measuring the torsional wind-up experienced when the
coupling is exposed to torque by comparing the relative
circumferential positions of different locations along the
couplings axis.
The most practical way of measuring the coupling torsional
wind-up has been found to be the measurement of phase shift of
separate speed pickups mounted along the couplings length. In
this respect, torsional deflection torquemeters have become
synonymous with phase shift torquemeters (Figure 4).
Each variation of the phase shift measuring torsional
defection type torquemeter senses the relative positions of
opposite ends of the coupling using a pair of toothed flanges
which are made as an integral part of the couplings spacer.
The sensing devices are stationary, and work on the same
principle as a typical speed pick-up - where the flux field
around the pick-ups are changed every time a (steel) tooth on
the rotating coupling passes it. By monitoring the phase
relationship between toothed wheels affixed to each end of the
coupling, an indication of coupling twist is obtained.
TO PHASEMETER
. circumferential position of teeth induces a sinusoidal voltage in
Figure 4. Basic Operating Principle Of Torsional
Deflection Type Torquemeters
All phase shift type torquemeter couplings must find a
solution to the problem that vertical and horizontal movements
of the rotating coupling relative to the non-rotating pick-ups
will also produce a phase shift. The method of distinguishing
between torsion induced phase shift and those caused by these
lateral movements form the basic differences of torsional
deflection torquemeter systems.
For phase shift type torquemeters the voltage signals
sensed by the pick-ups are typically sent to the control room
where the signals are then processed. As with strain gage type
torquemeters, the output of the processing unit is also typically
made available as an industry standard analogue signal for
connection to the users data recorders or PLCs. All phase
shift torsional deflection torquemeters measure coupling rpm as
a bi-product of torque determination, so the torque and speed
signals are typically multiplied for a direct readout of power.
Dual Channel Phase Shift System
One manufacturer deciphers genuine twist induced phase
shifts from false laterally induced ones by providing a version
of pick-up which senses the position of the toothed wheels
around many points along the circumference of the coupling.
Rather than supplying an array of circumferentially mounted
speed pick-ups, each pick-up takes the form of an internally
toothed ring. Each of these two pick-up rings (Figure 5)
completely surrounds each rotating toothed wheel of the
coupling. Each pick-up ring consists of circumferentially
wound coils, which are then energized to create a toroidal flux
path.
Figure 5. Pick-Up Rings Used in 2 Channel System
As the coupling lifts on one end relative to that pick up
ring, the early signal on the front side of the coupling is
canceled by an equally late signal at the back side of the
same toothed wheel. In this way, as the average
the driving side pick-up ring (Channel A). Similarly, the
average circumferential position of teeth induces a sinusoidal
voltage in the driven side pick-up ring (Channel B). The
processing unit receives these sinusoidal voltages of the
averaged position of the driving wheel on channel A and
compares it to the similarly averaged location of the driven
wheel on channel B. The difference in phase of these two
channels is a direct measure of torsional wind-up deflection.
By multiplying this torsional deflection by the predetermined
coupling stiffness (measured during initial laboratory
calibration), the processing unit is then able to determine the
torque present in the coupling.
4 Courtsey of Kop-Flex Emerson Power Transmision Co
3 Channel Phase Shift System
Another method of deciphering twist induced phase
shifts from false laterally induced ones is to move the two
toothed wheels of the coupling closer together to minimize the
errors caused by lateral movements of the coupling relative to
its housing. Simply shortening the length of tube between
toothed wheels would not work, because it would result in a
proportionally lesser amount of torsional twist
Instead, the (first) toothed wheel which is to be moved
closer to the (second) toothed wheel is still affixed to the
coupling at the same axial location as the two channel system -
but the manufacturer adds a non-torque carrying reference
sleeve (Figure 6). The function of this reference sleeve is
simply to translate the apparent location of the (first) toothed
wheel, moving it closer to its paired (second) toothed wheel,
without moving the location of the attachment point to the
torsion tube underneath the sleeve.
SECOND FIRST
Figure 6. Basic Operating Principle of Reference Sleeve
Used in 3 Channel System
In this arrangement, the reference tube is attached to the
torque-carrying torsion tube at one end only. The attachment
side is at the position of where the first wheel would have been
if it had not moved over to the other side. The reference sleeve
carries the first wheel to the other end of the coupling by
cantilevering off its fixed end. By moving the axial location of
the first toothed wheel closer to that of the second toothed
wheel, the error of false laterally induced phase displacement
is reduced, but since there is still some distance (x) between
these toothed wheels, the error has not been eliminated.
This now smaller error is canceled by adding a third
wheel (Figure 7). The third wheel is positioned exactly the
same distance (x) from the second wheel, but on the opposite
side of it. It is directly attached to the second wheel, so that any
phase shift seen between the second and third wheel is
attributable only to lateral movements (not from torque). In this
way, as the coupling moves laterally within the housing -
earliness of the signals for the first wheel can be canceled by
subtracting the amount of the lateness of the third
(compensating) wheel.
SECOND FIRST
THIRD
Figure 7. 3 Channel Phase Shift System With 3rd
Compensating Pick-Up
The sensor design is basically that of a standard variable
reluctance proximity probe - consisting essentially of a wire coil
wound around a permanent magnet. As with the two channel
system, as each tooth passes through the flux field of its mating
sensor, the processing unit receives a sinusoidal voltage of the
average position of the driving (first) wheel on Channel A.
Similarly, a sinusoidal voltage is received by the processing unit
for the average position of the driven (second) wheel on
Channel B. Likewise, a sinusoidal voltage is received by the
processing unit on Channel C for the average tooth position of
the compensating reference (third) wheel.
The processing unit (typically located in the control room)
receives these sinusoidal voltages of the average location of the
three wheels on channels A, B, and C. The phase difference
between channel B and channel C is purely the laterally induced
phase shift error. Knowing this, the processing unit subtracts
this error phase difference from the phase difference between
channel A and channel B. This corrected phase shift is then
the value associated with torsional wind-up only.
As with the two channel system, the torsional deflection is
multiplied by the predetermined coupling stiffness (measured
during initial laboratory calibration), where the result is the
torque present in the coupling.
Single Channel Phase Shift System
The single channel system is basically a modification to the
three channel system. The one channel system combines all
three channels into one channel - giving rise to the term
monopole, often referred to with this type of phase shift
5 Courtsey of Kop-Flex Emerson Power Transmision Co
torsional wind-up torquemeter.
The way in which this is done is as follows. Like the three
channel system, the one channel system moves the driving
wheel of Channel A closer to the driven wheel of Channel B.
Recognizing that the closer the two wheels can be made, the
lesser the error associated with laterally induced phase shift -
the wheels are moved so close together that they become
intermeshed (Figure 8). This feature results in the elimination of
the need for the other two channels (B & C), because the
intermeshed teeth result in sensing the tooth locations in a
single plane. As the coupling moves laterally within its
housing, relative lateral displacement of the two wheels is
therefore eliminated. Wherever the coupling moves relative to
the sensor, both sets of intermeshed teeth are in the same lateral
position.
TEETH OVERLAP
SENSOR
Figure 8. Intermeshed Teeth of Single Channel System
With the single channel system, instead of determining the
phase shift between two separate speed pick-ups, the phase
shift is determined between consecutive pulses of a single speed
pick-up. To illustrate, imagine that the two wheels are
intermeshed so that the teeth of the driving wheel are placed
equidistant between the adjacent teeth of the driven wheel. As
the coupling is rotated under no torque, the single sensing pick-
up will receive equally spaced increments of voltage pulsations
as each tooth on both wheels pass through its flux field. As the
coupling experiences torque, the driving teeth will move to an
earlier location. The resulting signal to the sensor will be
non-equally spaced increments of voltage. By comparing the
relative time lag of these non-equally spaced voltage pulsations
to the equally spaced ones (which were seen when the coupling
was spun under no torque), a phase shift of consecutive
pulsations is established.
This sinusoidal voltage signal is sent to the processing unit
(typically located in the control room) whereby the relative
phase shift between consecutive pulses on a single channel is a
direct measurement of torsional wind-up deflection - without
the need for correction for laterally induced errors. As with the
two channel system, and the three channel system, the torsional
deflection is multiplied by the predetermined coupling stiffness
(measured during initial laboratory calibration), where the result
is the torque present in the coupling. Since the coupling rpm is
once again an indirect result of the frequency of the sinusoidal
signal, the processing unit is also able to directly output the
power being transmitted by the coupling.
The manufacturer of the monopole Single Channel System
has also eliminated the radial support bearing that was often
found at the free end of 3 Channel Systems. The purpose of this
radial bearing was to eliminate any projected offset of the
cantilevered wheel if the torsion tube underneath were to bow
from misalignment. In such cases, a bearing was often used to
restrain the free end from being allowed to become laterally
displaced.
With the current Single Channel System, a second
independent torque reading is taken on the same set of toothed
wheels to provide redundancy for increased system reliability.
By choosing the location of the redundant probe to be on the
opposite side (rear instead of front) of the tooth set, the need for
a support bearing at the free end has also been eliminated. (See
Figure 9). If the torsion tube underneath the reference sleeve
bows under misalignment, any lifting of the free end of the
cantilevered wheel will produce an early signal between
consecutive pulses in the front of the coupling, with an equally
late signal at the rear. Any difference found between these two
independent phase shift measurements (remembering that each
phase shift is determined via consecutive pulses on a single
channel) is then attributable to a laterally induced false phase
shift. The average phase shift of these two readings is the result
of torsional wind-up alone.
INTERMESHED
Driving PICKUP TEETH
Intermeshed Pickup Teeth Driven
Machine
Machine
FLEXIBLE
COUPLING
Flexible Coupling
MONOPOLE
Monopole Sensors
SENSOR
EPU
Figure 9. Single Channel System with Second (Redundant)
Pick-Up
A New Torquemeter Development
Higher operating speeds, shorter shaft separations, and
lighter coupling weight requirements have led to a new
6 Courtsey of Kop-Flex Emerson Power Transmision Co
variation of the traditional cantilevered tube style torquemeter
coupling of the single channel system. This latest development
in the evolution of the torsional deflection type torquemeters is
a powerwheel torquemeter cartridge, depicted in Figure 10.
With this design all the required twist for measurement takes
place in a compact, specially profiled double wheel, where teeth
mounted to the outer diameter of one wheel move relative to the
teeth mounted on the outer diameter of a mating wheel, when
subjected to a torsional load. This movement is detected by the
same sensor as in the traditional cantilevered tube design of the
single channel system.
But, instead of a longer tubular section deflecting
torsionally under load, the torsional wind-up takes place within
profiled spokes in each wheel. The pair of spoked wheels are
connected together via a short tubular section at their inner
diameter. The speed pick-up teeth are located at the outside
diameter of each wheel.
TEETH
SPOKE
WHEEL 2
TUBE
WHEEL 1
Figure 10. Powerwheel U.S. Patent 5,969,269
Why A Powerwheel
The Turbomachinery Industry is constantly improving the
performance of high speed compressors, gearboxes, and
turbines, and this in turn is placing demands on coupling
suppliers to design couplings that can handle higher loads at
higher speeds. In addition, these couplings must weigh less (so
that the overhung moments on the connected equipment shafts
are low enough to keep the equipment from being highly
sensitive and having a tendency to vibrate excessively).
Furthermore, the couplings must be soft torsionally, to keep
the equipment trains from torsional resonant frequency
interferences, but stiff laterally, especially in longer shaft
separations. (A longer shaft separation is generally thought of as
greater than the minimum 18 inches specified by API 671 see
reference). A coupling must be stiff enough laterally to keep the
fundamental critical speed of the coupling away from the
running speed.
So, in many newer applications, a coupling must have low
weight, high lateral stiffness, low torsional stiffness, high torque
capacity, and must operate at high speeds. (They must also be
able to be manufactured quickly and inexpensively !!)
Add to all this the requirement to measure torque - another
reason to be soft torsionally - without significantly increasing
the weight, and you have an idea why a design like the
Powerwheel is needed. It is lightweight, soft torsionally and
stiff laterally. It easily handles the applications that are marginal
for 1 and 2 channel phase shift designs.
Figure 11 is a depiction of a coupling in the initial tough
application that prompted the development of the powerwheel;
an application where it was difficult to design the standard
(non-torquemeter) coupling without special features. This
coupling goes between a 7550 hp gas turbine and a centrifugal
compressor operating at 13600 rpm. Note the reduced moment
(low overhung moment) hub design on the compressor end, and
the larger than normal spacer tube diameters designed to keep
the coupling laterally stiff in this long" shaft separation (31.5
inches) application.
DRIV EN UNIT
DRIV ING UNIT
Figure 11. Non-instrumented coupling
In Figure 12 is the first attempted design for a torquemeter
coupling, a single Channel cantilevered tube-twist type. This
design led to unacceptable bearing vibration during a high
speed test. The critical speed of the coupling itself, compared
to operating speed, was too low. A redesign of this cantilevered
concept, Figure 13, where cantilever length was significantly
reduced, also had unacceptable coupling critical speed margins.
7 Courtsey of Kop-Flex Emerson Power Transmision Co
 Note also that this stiffness changes with speed and
PROBE
temperature).

DEVELOPMENT OF A HIGH SPEED TORQUEMETER

MODAL ANALYSIS SUMMARY
LONG SPAN (31.60-31.67" DBSE)*
FREQUENCY (CPM)(Thousands)
28
# NON-INSTRUMENTED (REF.)
NON- SLEEVE DESIGN
TWIST TUBE WELDED # #
TEETH INSTRUMENTED
26 ) POWERWHEEL-STYLE DESIGN #
# ) )
24
GAS TURBINE COMPRESSOR # )
22 ) POWERWHEEL
Figure 12. Cantilevered tube-twist torquemeter
20 )

18 #


16
) SHORT TUBE
PROBE TWIST TUBE 14,300 RPM
14

12
1.0E+05 1.0E+06 1.0E+07
END SUPPORT STIFFNESS (LB/IN)

Figure 15. Modal analysis summary

At a typical overall end support stiffness of 1,000,000 lb/in,
TEETH the powerwheel coupling critical speed is 23800 cpm compared
to the "short tube" cantilever design value of 17500 cpm.
Figure 13. Short-tube cantilevered tube twist torquemeter Figures 16 and 17 show the relative mode shapes and deflection
magnitude between the short-tube cantilevered design and the
powerwheel.
In addition, the powerwheel has other advantages. It is
POWERWHEEL compact axially, so that long lengths of tube-twist are not
TEETH required. Short shaft separations can be accomodated, and
retrofits can be easily fitted to existing applications. Since
initial development, the powerwheel spoke versatility has
allowed for designs to measure in the range of 3,000 in-lb of
torque on the low end, and 1,300,000 in-lb at the upper end.

How It Was Done

PROBE
Figure 14. Powerwheel torquemeter
What worked, and gave acceptable margins between the
coupling critical speed and the operating speed was the
powerwheel design, Figure 14. The graph in Figure 15 gives an
idea of the superior frequency margin with this design, as a
function of the coupling end support stiffness. (Note that the
equipment shafting and geometry, in addition to the fluid film
journal bearing design, all contribute to the end support
stiffness, and, therefore, the actual coupling critical frequency.
To accomplish this compact design, Finite Element Analysis
(FEA) and proprietary numerical methods were used to design
the spokes of wheels. They had to be strong enough to handle
the stresses due to torque and centrifugal effects, and soft
enough to twist and deflect enough to generate a good electrical
signal. Experience with designing diaphragm type couplings,
where metal is deflected in the axial direction to accommodate
misalignment, and is loaded in the circumferential direction by
torque, benefited in the design of the powerwheel, where the
spokes are deflected circumferentially by torque.

8 Courtsey of Kop-Flex Emerson Power Transmision Co

 mils FIRST LATERAL NATURAL FREQUENCY: 17,600 CPM, CLOSE-UP OF COMBINED VON MISES STRESSES @ 45,000 LB.IN. & 11,000 RPM
BASED ON 1E6 LB/IN END SUPPORT STIFFNESS

SPOKE

PSI

Figure 16. Short tube analysis

mils FIRST LATERAL NATURAL FREQUENCY: 23,200 CPM,

BASED ON 1E6 LB/IN END SUPPORT STIFFNESS

Figure 18. Wheel stress analysis

POWERWHEEL SYSTEM STUDY

MEASURED STRESS DATA VS. THEORY (ALGOR FEA)
Stress (KSI)
55
)
50 )
)
45 Long Spoke Theory )
Figure 17. Powerwheel analysis
40 )
)
35 )
)
30
Analysis and Testing ) Short Spoke Theory
25 )
)
To determine the stresses from torque and centrifugal 20 )
effects, an FEA model was necessary for this complex 15 )
) ) Long LegMEASURED
Test
geometry. The spokes, the short tubes between the wheels, and 10 ) Short Leg Test

the overhanging signal teeth, which deflected at the largest 5 )

diameter from speed, were the critical areas. All were modeled )

0
)
with FEA and the resultant stresses under load were compared 0 5 10 15 20 25 30 35 40 45
to actual strain gage measurements. Figure 18 shows the
Torque (lb-in) Thousands
maximum combined stresses in the wheel from torque and
speed, while Figure 19 shows a graph of strain gage results
compared to FEA predicted results. The correlation is excellent
(within 5%). Figure 19. Stress - measured vs. theory

Stiffness Calibration
Using calibrated weights precisely positioned at the end of a
long torque arm, an optical encoder is used to monitor the tooth
The most important calibration for the torquemeter
movement as a function of applied load (torque). The
coupling, and therefore the powerwheel is the torsional stiffness
deflections match closely the deflections predicted by FEA
calibration. The principal that metal will deflect by a known
analysis (Figure 20). Even more, the wheel deflections prove to
amount given a known load (in its elastic range) is crucial to
be linear with torque.
torquemeter design.

9 Courtsey of Kop-Flex Emerson Power Transmision Co

 position after repeated cycles of speed and torque load. This is
TORSIONAL STIFFNESS TESTS OF THREE PROTOTYPES important especially considering the small amount of movement
VS. THEORY (ALGOR FEA) that is being detected by the probes 0.01 mils.
Angle of Twist (deg) After repeated dynamic tests with and without torque
0.7 applied, the rings remain in the same place and the tooth
#
# readings are repeatable (Figure 22). Note that since the system
0.6 # is so repeatable, even if the actual reading is different by a
#
# & significant amount than what might be calculated by heat
0.5 &
# & balance, using the baseline data, significant trends can be
# &
& monitored. Predictive maintenance can occur.
0.4 # &
# & Note also that the probe is mounted in a stationary guard
&
# & which can move relative to the coupling in operation. The guard
0.3 # &
&
# is mounted to the connected machine housings, the coupling to
# &
&
0.2 # &
& the connected machine shafting. As a result, considering the fact

# & that at higher speeds the overhang teeth will deflect outward,
&
0.1 # & Steel Test # Aluminum Test &TitaniumTest errors will result in the readings, depending on the probe
# &
Steel FEA Aluminum FEA TitaniumFEA
#& & position in relation to the teeth.
0
&
0 5 10 15 20 25 30 35 40 45 50 55
Torque (lb-in) Thousands
ZERO TORQUE OFFSET READINGS AFTER STATIC AND DYNAMIC TESTING
(RELIABILITY OF THE SHRINK-FIT RINGS ASSEMBLY)
Figure 20. Stiffness - measured vs. theory ERROR (% OF FULL SCALE TORQUE)
1.00
#AFTER STATIC CALIBRATION @ 45,000 LB.IN
0.80 "AFTER HIGH TORQUE DYNAMIC TESTS
Finally, it is known that the modulus of different metal !AFTER HIGH SPEED DYNAMIC TESTS
0.60
materials change slightly with temperature, which could affect
the output of the torquemeter when trying to detect small 0.40
TEST SETUP REPEATABILITY RANGE: +/- 0.1%
changes in machine torque output (less than 1%). An RTD 0.20
device is inserted in the sensor to detect changes in temperature " " " " " " " " " "
0.00 # # # # # " # # # #
to adjust the torque or power readings. Lab testing, where the # ! ! !
#
! ! ! ! !
-0.20 !
! !
powerwheel was heated to different steady state temperatures in
the expected operating ranges, correlated the effects of -0.40
temperature with torquemeter output.
-0.60

-0.80

-1.00
10,000 10,500 11,000 11,500 12,000 12,500 13,000 13,500 14,000 14,500 15,000
TEST SPEED (RPM)

Figure 22. Shrink ring reliability

Testing proved that at any possible probe movements, the

reading errors were minimal. The probe can move axially along
the teeth at the same time it moves toward or away from the
teeth without significant error. (Figure 23).

Modal Analysis

Since acceptable lateral critical speed margin was the main

Figure 21. Gear rings impetus for the powerwheel design, it was important to be able
Gear Ring Loads and Tooth Deflections to accurately predict these margins for different cases. The
Because of its unique design, the powerwheel gear rings method of modeling (FEA, Figures 16, 17) that was developed
are shrunk fit onto the powerwheels (Figure 21). This raises the needed to be correlated with actual measurements. These
question of whether or not the ring will move from its original

10 Courtsey of Kop-Flex Emerson Power Transmision Co

measurements were taken in the lab, and the correlation for four case scenario. A typical error budget is shown in Figure 25
(4) different designs is shown in Figure 24. These errors are combined to give an rss Total System Error.
The powerwheel system is guaranteed to have an error rate of
ZERO TORQUE OFFSET ERRORS DUE TO AXIAL MISALIGNMENT AND SPEED
less than 1% rss.
(INCLUDING RADIAL TOOTH GROWTH EFFECTS)
ERROR (% OF FULL SCALE TORQUE)
1.00
Error Computation
0.80 # ZERO AXIAL Error Expressed as Error Component
0.60
0.050" AXIAL, INBOARD
(in-lbs) (El. deg.) (mils) (%)
) 0.050" AXIAL, OUTBOARD
0.40 000 66 0.0081 0.048 .146% Speed Comp. Error
000 82 0.0102 0.030 .182% Radial Gap Error (.015")
0.20
9.7 123 0.0153 0.090 .273% Axial Movement Error (.100")
0.00 )
# #) #) #) )
# #)
#
)
#) 650 45 0.0056 0.033 .100% Mechanical Calibration (optical encoder)
#) #) #)
24 0.0030 0.018 .054% Zero Torque Setting Repeatability
-0.20
4V 48 0.0060 0.035 .107% Setup Repeatability
-0.40
00 54 0.0067 0.040 .120% Temperature Compensation (200 C)
-0.60 000 77 0.0096 0.057 .172% Sensor Variability (swapping)
700 77 0.0096 0.057 .172% Electronic System Accuracy
-0.80
19 48 0.0060 0.035 .107% Electronic System Variability (swapping)
-1.00
45 0.0056 0.033 .100% Cabling Effects (300 feet)
10,000 10,500 11,000 11,500 12,000 12,500 13,000 13,500 14,000 14,500 15,000
6
TEST SPEED (RPM)
00
12 224 0.028 0.164 .498% Total System Error (RSS)
Figure 23. Errors due to misalignment and speed
500
466

TEST C ON DITION S THEORY (ALGOR MODAL FEA)

596
S UP PORT METHOD "A"* S UP PORT METHOD "B"** "K" = 250,000 LB/ IN "K" = 1.0 E6 LB/IN "K 3.0
ME AS URED LNF (CPM ) ME AS URED LNF (CPM) LNF (CPM ) LNF (CPM )

22,000 i 23,500 22,600 25,800 Figure 25. Error budget

E EL 19,000 22,050 20,100 23,500

EEL 19,650 ii 20,400 19,620 22,080 Dynamic Testing

EEL 20,400 21,750 19,860
Figure 24. LNF - measured vs. theory
The top line in Figure 24 (i) is the non-instrumented
coupling, while the next three (ii) are differently designed
powerwheels. When conducting the LNF test, the coupling was
supported in two different ways, Method A to simulate a
softer support stiffness, and Method B, stiffer. The range of
actual installed machine support stiffness is 250,000 lb/in to
1,000,000 lb/in, represented in the last two columns. The
correlation is very acceptable.
Error Budget
One of the major requirements for the torquemeter is
accuracy. An overall error rate of more than 1% can mean
significant long term cost. There are many contributors to this
error rate, and they are all summed and added to get a worst
23,200
Modeling was done to simulate to combined stresses due to
speed and torque, then comparisons were made with lab tests.
One major test was a back to back test where a proven 1
Channel tube twist meter was direct connected to a powerwheel
torquemeter and loaded with torque at high speeds (Figure 26).
Each meter had a separate probe and an independent electronic
system. Correlation was within 0.3% of full scale torque, as
shown in Figure 27.
Initial Continuous Field Units Experience
For reference, Figure 28 depicts an installed probe in the
bottom half of a coupling guard (with the top half removed), of
a cantilevered tube twist single channel phase shift system;
Figure 29 shows a probe installed in the top half of a guard
thats been buttoned up.
Similarly, two initial powerwheel units were installed each
between identical gas turbine / centrifugal compressor gas
transmission trains , 7500 hp at 14000 rpm. At this site, in 1998,
the system was first successfully used to prove the compressors,
as compared to factory tests. Now it is used to monitor the
installation for performance degradation.

11 Courtsey of Kop-Flex Emerson Power Transmision Co

 BACK-TO-BACK WELDED SLEEVE - POWERWHEEL TORQUEMETERS
(CLOSE-UP VIEW, HIGH SPEED TESTING)

SENSORS

WELDED SLEEVE
POWERWHEEL
Figure 28. View inside coupling guard of installed unit

Figure 26. Back-to-back testing

FIRST SENSOR (0) MOUNTED IN COUPLING GUARD.

POWERLIGN SYSTEM ACCURACY

TUBE TWIST VS POWERWHEEL
POWERLIGNSYSTEMACCURACY
HIGH
WeldedSleeve TORQUE
vs. TitaniumPowerwheelTESTING
(twosensors), HighTorqueTesting
50,000 1.00

45,000 0.80

40,000 0.60

35,000 0.40
% Difference of Full Scale Torque

30,000 0.20
Torque (in-lb)

25,000 % DIFFERENCE 0.00

20,000 (WHEEL VS TUBE) -0.20

WeldedSleeveTorque
15,000 -0.40
Powerwheel Torque
%difference=(Pwhl - Slv)

10,000 Full scale=45,000in-lb -0.60

Test Speed=1800RPM
FULL SCALE = 45000
WeldedIN-LB
Sleeve=5.51el. deg.
5,000 TEST SPEED = 1800 RPM
Titanium pwrwhl =5.47el. deg. -0.80 Figure 29. Fully installed unit - probe junction box mount

0 -1.00
9:55:00 9:56:00 9:57:00 9:58:00
Time

Figure 27. Back-to-back measurement differences CONCLUSIONS

1) Torquemeters are becoming a more important tool for

turbomachinery predictive maintenance.

2) The 1 channel phase shift powerwheel system is a viable

option for difficult high speed application torque monitoring.

12 Courtsey of Kop-Flex Emerson Power Transmision Co

ACKNOWLEDGMENTS

The authors would like to acknowledge Mr. Jossef

Zilberman and Mr. Robert Munyon who came up with the
idea and did most of the development work on the
powerwheel.

REFERENCES

Kurz, R., Brun, K., and Legrand, D., 1999, Field Performance
Testing of Gas Turbine Driven Centrifugal Compressors,
Proceedings of the 28th Turbomachinery Symposium,
Turbomachinery Laboratory, Texas A&M University,
College Station, Texas, pp.216-220.

Wilcox, E., 1999, Practical Methods For Field Performance

Testing of Centrifugal Compressors, Proceedings of the
28th Turbomachinery Symposium, Turbomachinery
Laboratory, Texas A&M University, College Station, Texas,
p. 168.

API Standard 671, 1998, Special Purpose Couplings For

Refinery Service, Third Edition, p.3.

ASME PTC 10, 1997, Performance Test Code on Compressors

and Exhausters, American Society of Mechanical
Engineers, New York, New York.

R.D. and J.D. van Millingen, 1991, Phase Shift Torquemeters

for Gas Turbine Development and Monitoring, ASME
Paper 91-GT-189, p.1.

13 Courtsey of Kop-Flex Emerson Power Transmision Co