You are on page 1of 28

A STRAIN GAUGE TORSION}IETER FOR SHIP SHAFT SYSTEMS

by

H.B. Boyle, C.Eng., A.M.I.Mech.E., A.M.R.I.N.A.

Introduction
The measurement of propulsive power is required at least once in the life of
a ship, and quite often the measurement of power or torque is made regularly as
part of a general study of performance. This paper describes the experience
gained on one type of torsiorimeter, namely that using wire or foil resistance
strain gauges.
In 1957, Ship Division of the National Physical Laboratory started upon a
research programme involving the need for a torsionmeter to measure torque on small
diameter (k in) shafts. Because of the small diameter involved, (making most
other systems impracticable) together with the general strain gauge experience of
the Division, it was decided to use a strain gauge system. Since that date a
number of shaft systems have been gauged, all the while building up a considerable
background knowledge. Critical examination of the results has led the author to
believe that a strain gauge torsionmeter is as good as any system for all practical
applications and has a number of advantages over most.

Choosing a Torsiônméter System


Ship torsionmeters, almost without exception, use the propeller shaft as the
flexural element; the possibility of using a special shaft section, preferably
weakened to obtain higher stress concentration or larger deformation is almost
non-existent. Even with this limitation many measuring systems have been proposed
and quite a number are in regular use. (References 1 and 2).
The initial choice of torsionxneter may depend upon a number of factors such as
cost, commercial availability, or convenience of use. Thus when choosing a system
it does not necessarily follow that the right system is chosen. For this reason
there must be a critical examination, at regular intervals, to ensure that the
best engineering methods are being used.
From simple theoretical considerations there are only two basic systems
available if it is assumed that the shaft is to be the transducer element of any
torsionmeter system, and some form of mechanical deformation is to be measured.
Consider the standard engineering relationship.
-2-

Jr
T q GQ
1
where T
J
=

=
Torque
Polar moment of inertia
q = Sh.ar stress
r = Radius
G = Shear modulus
Q = Angle of twist
1 = Length
this may be rewritten as

qJ GQJ
T
r

Since shear stress cannot be measured directly, it is necessary to measure the


principal strains resulting from the shear stress, hence

2GeJ GQJ
T where e = direct strain.
r 1

Assiuning J,l and r are known constants, then a chosen system must measure
either:-
Method (1) GQ or Method (2) Ge

Therefore from theoretical considerations, each system requires the determination


of two unknowns. (G is normally measured by ultrasonic methods - Ref 2).
Besides these theoretical considerations, there are a number of practical
problems to be solved. These may be divided into three sections, namely:-
Choice of measuring transducer.
Signal transference from shaft to measuring system.
Suitable accurate measurement and read-out.
The electrical measurement system has almost suppressed the use of direct optical
or mechanical systems, probably for reasons of convenience, remote read-out and
the advantage of electrical magnification being the most important.
Let us then consider the engineering problems for a typical electrical
torsionmeter system.

(1) Choice of measuring transducer


Consider the two methods mentioned above:-
Method 1 (GQ) Normafly such a system does not measure the angle Q, but a linear
displacement having an electrical signal proportional to tan . In order that
this measured displacement is large, the two points which define 1 should be as
widely separated as possible. If 1 is very large it is necessary to measure the
displacement from stations mounted on the ships structure. Unfortunately a ship
cannot be considered a rigid structure and with such a system large errors are
'-3

likely to occur due to hull flexing. There is the added problem of shaft
couplings for which the stiffness may be difficult to determine. If the displace-
ment is to be measured with reference to two points on the shaft, then 1 will
probably be limited to a maximum of 3 ft. This is due to the conflicting require-
ments of producing a light easily portable system with sufficient rigidity not to
be influenced by shaft rotation. The whole system would normally be calibrated on
a dummy shaft subjected to known angles 0 and then clamped to the ship's shaft.
The method of clamping which seems to be preferred is by means of half rings
bolted together. Although this system gives rise to doubts as to the effective
clamping position, and hence the length 1, it does provide a very rigid method
of attachment.
Method (2) (Ge) The torsional shear stress produces principal strains on helices
which are at +5° to the shaft axis. It is therefore necessary to attach some form
of extensometer or strain measuring device. Since the measurement of strain is
along a curved path, the conventional mechanical extensometer is unsuitable and
the most obvious choice for such a measurement is the resistance strain gauge
Accurate positioning of the gauge is important, but if strain is being measured
the length over which the measurement is made is not important. However the use
of the strain gauge involves one further variable quantity, namely gauge factor,
which is the ratio of electrical to mechanical strain.

Electrical Signal Transference


The electrical output from a torsionmeter must be transferred from the
rotating shaft to a recording station. There are a number of methods available
including such sophistication as telemetering systems. Such systems are
unnecessary 'complications for shaft torque measurements when slip rings using
silver rings and silver/graphite brushes ar so readily available. Experience has
shown that this type of transfer unit is re] iable and does not introduce extraneous
signals if used correctly. The quality of signals which may be transmitted is
shown in Fig.1. Here the shaft skin strain is of the order ±ij- microstrain
x o_6 inch per inch strain). The major fluctuations are due to propeller
force excitation and only the superimposed small amplitude, higher frequency ripple
can be attributed to extraneous signals.

Electrical Measurement
The mechanical. to electrical signal conversion may take a variety of forms,
the most common being voltage, frequency or phase dependent. Although the latter
two methods may appear to be preferable in that the signal amplitude is unimportant,
modern electronic technology has made possible the accurate measurement of
extremely small voltages.
14_

The Strain Gauge System


The Difficulties
In any strain gauge system whether for a ship torsionmeter, or for any other
purpose, there are a number of difficulties to overcome. These difficulties are
cormnon to all gauge installations so it is proposed to deal only with those that
require special attention in the construction of a ship torsionxneter. These may
be listed as gauge positioning, temperature effects, determination of gauge factor,
gauge bonding and sensitivity to other strain directions.
The correct positioning of a gauge depends on two factors, correct 'marking
out' on the shaft of the position lines and the bonding of the gauges to these
lines. Even with quite simple methods of marking out it will be difficult not to
achieve an accuracy of at least 10 and, with care, an accuracy of better than
1/100 is possible. (Yig.2 shows two lines at 10 angle of separation). The
resulting error would be .06% for 10 positional misalignment.
Gauge positioning can only be satisfactorily achieved with gauges where the
metal grid of the gauge is clearly visible and the gauge has a transparent backing;
most foil gauges are admirable for this purpose. The gauges can now be positioned
using a portion of the grid to coincide with the markings on the shaft. If gauges
of sufficient size are used it is reasonable to get a system which is correctly
positioned to approximately ± 1/10°.
Many wire or foil resistance gauges exhibit temperature strain sensitivity;
the use of 1 gauges in a torsionmeter helps to reduce this problem but .does not
eliminate it. An elegant solution is to use a gauge where the gauge temperature
sensitivity matches and cancels the effects due to the strain induced into the
gauge by the shaft thermal expansion. Such a system can give excellent
temperature stability. On one such small laboratory installation which was tested
a change of approximately 60 deg C induced an apparent strain of only microstrain.
This type of commercially produced gauge is again a foil gauge.
Determination of gauge factor is achieved only by measurement. it is the
accuracy of this measurement which has the most significant effect on the accuracy
of the system. (A simple device, as shown in Fig.3 enables the gauge factor to be
derived). At one time it was felt that gauges should be bonded to a test beam at
the same time as those bonded to the shaft, since it was thought that the gauge
factor might be influenced by environmental conditions. However experience has
shown that this is unnecessary.
The modern manufacturer carries out quite exhaustive tests to establish the
gauge factor of a batch of gauges, (Ref.3) and as long as the manufacturer's
recommended bonding techniques are foflowed, this gauge factor may be used without
recourse to any checking.
5

The arrangement of the gauges to measure torque is theoretically insensitive


to the two other main sources of strain, i.e. shaft compression due to thrust and
shaft bending due to shalt weight or misalignment of shaft bearings.
In practice the thrust sensitivity is small if good quality matched pairs of
gauges are used. A typical shaft designed to carry a torsional load of 12,000 lbf ft
resulting in 250 microstrain gave, when loaded to 10,000 lbf thrust, an output of
less than j microstrain.
Although the shaft skin strain due to bending may be large compared to that
for thrust, errors are only likely to occur in establishing the torque zero when
the shaft is at rest since the bending strain will be cyclic with the shaft
rotating and thus the variation of signal may be averaged. (This is usually
achieved electrically by using a recording device having a low frequency response
indication). Errors in determination of the zero position may be obviated by
averaging the readings obtained with the shaft slowly rotating. This technique
also reduces any residual torsion in the shaft between bearings.

Discussion of Equipment and Techniques


The four case histories presented here, summarised in Fig.k, have been picked
as each high-lights some particular aspect of the overall problem. In all cases the
measuring system was identical, and is described below.

1. Neasuring Equipment
This was a strain indicator of commercial make using a carrier system of strain
gauge bridge excitation, a 5 volt supply having a frequency of 2,500 or 1,000 cycles
per second. The system operates either as a direct reading meter device or as a
null balance system. A sensitivity switch changes the value of the full scale
deflection of the meter in ranges of 30e, 300te, 1,000i.te and 10,000e steps. The
null balance is achieved from two switch positions of 1O,000e and 1,000e steps
plus a continuous dial adjustment totalling 2,000.ie marked in 10e intervals. This
latter dial can be read to better than ±4te. The strain steps referred to assume
one active strain gauge having a gauge factor of 2. The readings for a four-arm
torsional strain gauge bridge must be modified as follows:-

Reading 2
Strain = x
No. of Gauges Gauge factor
-6
The absolute accuracy claimed by the manufacturer for this measuring system is as
follows: -

Accuracy as Null Balance Device ±%


Accuracy as Direct Reading Device ±1%
(Taken from Manufacturer's General Catalogue 1966).

Experience with a number of these strain indicators has shown that the order of
repeatability and accuracy is certainly higher than quoted. Provision is also
made for a rectified signal to be available from two output terminals, voltage
being related to the full scale deflection of the meter. Internal calibration
control and drift monitoring facilities are also provided. During calibration the
null balance system is used, but during measurement it is more convenient to use a
combined technique whereby balance is achieved in 100te steps, the residual being
read directly from the meter.

2. Slip Rings
Both commercial and "homemade" slip rings have been used successfully, the
choice is one of convenience, the commercial ring being easier to fit, the home-
made or constructed ring being about a quarter of the cost. Both types of ring
mounted on a shaft insert are shown in Fig.5. Wherever these units have been used
they behaved well and have shown no detectable wear. All four cases referred to
in Fig.k used. commercial slip rings and in case III a constructed unit on one shalt
was used for comparison purposes. In case II, the rings and brushes were
subjected to an estimated 1,000,000 revolutions without any visible deterioration.
In cases II and III the same commercial rings were used, the difference in shaft
diameter being accommodated by semi-circular packing pieces.
In case III, because of the gauge position it was possible to take zero torque
readings both when the shaft was stationary 3nd rotating at 380 r.p.m. The
difference between the zeros, stationary and rotating was not measurable and must
therefore have been less than 1 microstrain. This "rotating" zero was recorded
during single engine trials, and the trailing propeller caused the disengaged shalt
to rotate.
In cases II, III and IV the rings, whilst in use1 were subjected to
unintentional contamination by various fluids. Case II was positioned under a
fueld pipe leak, case III having one slip ring assembly splattered with grease
from a gear coupling, and in case IV, the drive gear box developed a leak which
spilled oil over the slip ring whilst running. In no instance was there a
noticeable effect on the readings, there being no erratic change of signal level
nor change of shaft zero from the original clean state.
-7
3. Brush Gear
In all cases commercial brushes made of silver/graphite were used. These
brushes, two per ring, were supported on beryllium copper canti].ever springs. The
brushes were attached at the connection end to a common insulating block, arid then
angled out to the shaft. The brush contact points were thus at about 600
separation around the shaft. This arrangement made for very easy setting up and
removal. Positioning, to obtain the correct spring pressure, was not critical and
the unit could be disengaged whilst the shaft was rotating.

k. Strain Gauges
In case I the gauges used were paper-backed, grid wound iridium platinum wire
gauges. These were chosen for the high gauge factor (6). This trial was part of
a research into propeller excited vibration (Ref.k). Under these circumstances it
was more important to record the fluctuating strains with good signal to noise ratio
than to achieve a measurement of mean torque to a high order of accuracy. However,
as an AEI torsionmeter, fitted by the British Ship Research Association was
available, it was decided to compare results obtained from both meters. The type
of gauge used was very sensitive to temperature change and to reduce this effect
careful gauge resistance selection was necessary. Two torque bridges were fitted
in the same axial position, but separated by 90° in the radial direction. The slip
ring assembly was mounted over the gauge position. The gauges were bonded with
rapid setting cement. Due to the extreme cold, local heating was introduced by
means of a butane gas burner in the shafting compartment of the ship. Whilst this
raised the shaft temperature from 0°C to a more acceptable temperature (about 15°C),
the humidity became extremely high, any unprotected steel rusting rapidly. In
case II, paper-backed Karma wire grid wound gauges were used. These gauges were
bonded on to the shafts out in the open alongside the ship's slipway. Electric
heaters were installed near the gauge positions and the whole covered with a
tarpaulin. During the 2k hour curing time it rained, again giving extremely high
humidity conditions. The gauges were then waterproofed and calibrated, and the
shafts installed in the ship. Approximately 2 months elapsed between calibration
and use.
In cases III and 1V1 epoxy-backed, etched foil self-temperature-compensated
gauges were used and bonded under laboratory conditions.

5. Gauge Glue
These are listed in the case histories but special mention should be made
about Araldite Twin Pack. This seems to be an. ideal strain gauge glue for field
work. Experiments in the laboratory have shown it to be as good as the conventional
-8
epoxy strain gauge resins, and it does not seem to be too critical as to the ratios
of the two part mix. Small quantities can be mixed easily and it can be purchased
at any local ironmongers. One word of warning; difficulty has been experienced on
diameters below 1 in. Under these circumstances the gauged units experience very
undesirable hysteresis. It is suspected that this trouble is due to a thick
deposit of glue over the centre section of the gauges If this is so it probably
stems from the gauge bonding technique, where little or no bonding pressure is
applied.

6. Revolution Counting
For case I, in conjunction with the slip ring manufacturer, NPL devised a
method of pulse counting. This consisted of a segmented ring composed of alternative
silver and epoxy resin segments, the segments being incorporated into the slip ring
construction. The silver segments were joined together electrically during
manufacture and earthed to the shaft during installation. A capacitance probe was
then used as a pulse generator, the pulses being fed to an electronic counter via a
suitable amplifier system.
The system has proved to be very reliable. It has a number of advantages over
reflective light, photoelectric or inductive systems, since it is not affected by
oil or dirt and the pulse size is not related to the shaft rotational speed. In
case II, NPL were not required to measure power, only torque, and so no tachometer
was fittede
In case III, the high speed diesel engines were fitted with shaft pick-up
points suitable for hand-held mechanical tachometers. For this particular trial,
where the engines were speed governed, a hand-held tachometer was considered more
convenient and was used.
In case IV, the engines were supplied 'ith electronic counters.

Measurements and Results


Case I
Readings were taken of the outputs fron the two strain gauge bridges, the AEI
torsionmeter and the NPL shaft tachometer. These readings were taken over the
major part of the shaft revolution range, whilst the ship was proceeding along the
English Channel. The weater conditions were poor and this was reflected in the
general random scatter of results, probably due to ship motion which resulted in
continually changing shaft rotational speed. This was aggravated by aU readings
being taken in serial form; due to limitations of equipment then available it was
necessary to switch each strain gauge bridge, in turn, into a single strain
indicator. Since records of fluctuating strain were being taken there was, of
necessity, a considerable time lapse between measurements of the two bridges.
9

It must be stated that the experiment was not performed essentially to


compare torsionmeter accuracies but was part of an overall vibration study. The
results obtained from the strain gauge and AEI torsionmeters are shown compared in
torque
Fig.6. They have been plotted as versus shaft
(shaft rotational speed)
revolutions; the latter plotting provides an expanded vertical axis. The AEI
torsionmeter readings were converted into torque on the basis of a formula
provided by BSRA. The formula involved the use of a correction factor obtained by
calibration. This calibration is shown in Fig.7.

Case II
The shafts for this ship were calibrated by means of a very simple calibration
rig; the results of the calibration are shown in Fig.8.
To calibrate, each shaft in turn was laid on a concrete floor and restrained
at each end in the vertical direction by two metal straps grouted to the floor.
Single torque arms were bolted to each end of the shaft, through the normal shaft
coupling flanges. One arm was bolted to the floor, the other arm being attached at
its outer end to a tripod via a tension screw jack and a measuring hydrostatic
capsule (accuracy ±5 of full scale deflection). The torque applied was equivalent
to the capsule reading times the torque arm distance of 10 ft. To reduce the
errors due to friction, the rig was hit at the metal strap positions with a heavy
mallet at each load condition.

Case III
A simple calibration rig was constructed at NPL and the constructional details
are shown in Fig.9. The load measuring capsule was similar to that used in
Case II. The torsionmeter units were calibrated twice and used for ship trials on
three occasions over a period of eleven months. Note was taken of zero shift and
overall repeatability and these are discussed later. The calibration was made at
Ship Division, Feltham, whilst all trials work was carried out on the Solent.
Calibration results are shownin Fig.10.
Although no other torsionmeter was available, the engine manufacturer had
calibrated the engines on a brake at the factory, arid, during the trial, records
of r.p.m. and various engine parameters were taken. From this information
horsepower figures were derived. A comparison of manufacturer's and NPL values of

power are shown in Fig.11.

Case IV
In this series of trials the shipbuilder and NPL produced 12 shaft torsion-
meters, each shaft being fitted with two strain gauge bridges. The shafts were
calibrated in both the port and. starboard turning directions in the rig mentioned
- 10 -

for Case III. Three shafts selected at random, as far as the strain gauge bridges
were concerned, were then sent to the engine manufacturer to take part in engine
and gear box trials. This enabled a comparison to be made between a calibration
of a simple static type and results obtained froi a Heenan and Froude water brake
under running conditions. The results of these calibrations and trials are shown
in Figs.12 and 13.

Discussion of Results
Case I
The results, Fig.6, show good agreement between the two measuring systems. It
was to be expected that the two strain gauge bridges would be in better agreement
with each other than with the AEI torsionmeter, since they were measured at closer
intervals of time. However, there is no doubt that the best smooth line, as
defined by the strain gauge system, was confirmed by BSRA readings. The results of
the AEI torsionmeter static calibration are shown in Fig.7, and it is interesting
to note the change of calibration factor with respect to output reading; this
variation is approximately 3%. Although the readings were converted to torque by
use of this correction curve and calibration formula, it is preferable to have a
torsionnieter which has a more linear relationship between torque and output reading.
As will be seen in Cases II, III and IV, strain gauges meet this requirement.

Case II
The object of including the results of this trial is to show the possibilities
of calibration of smaller shafts by simple means. The calibrations in Fig.8 show
the repeatability that is possible on two separate shaft systems.
The general shape of the calibration curve is influenced by a number of
factors; friction within the calibration system, load capsule calibration errors,
and the inaccuracies caused by the displacement of the torque arm through a large
angle under load. This displacement was caused by the twisting of the shaft under
torsional loading and the flexibility of the torque arm. Since the jacking anchor

point was fixed, there was an effective change of length of the torque arm. (See

Fig.lk).
The friction losses almost certainly produce the steeply falling portion of
the curve in the lower part of the calibration, whereas the rising portion of the
curve is due mx4inly to the angular displacement error mentioned above.

Case III
This set of results (Figs.10 and ii) is excellent, both the repeatability
achieved over a period of months and the stability of the zero indicate that the
strain gauge system offers long term stability for torsionmeter systems. The zero
scatter was of the order of ±1 i/'+% of Full Scale Reading spread over a period of
8 months. This period involved two trials on separate ships, two calibrations in
the laboratory, and the use of two similar strain indicators. Using the shaft
diameters and gauge factors, the calculated values of modulus of rigidity of the
two shafts are 11.91 x 1O6 lbf/inch2 and 11.97 x lbf/inch2, which are
satisfactory values for steel, lying within the spread of modulus measurement made
by the BSRA (Reference 3). Again, there is the steeply falling curve at the lower
end of the calibration due to friction, but even accepting this possible source of
error the dynamometer has a linearity of better than ±1% over half of the torque
range and better than ±2% over four fifths of the range. It is now normal
practice to ignore the lower portion of the calibration curve and to establish the
calibration factor by extrapolation of the upper part of the calibration towards
zero. In this particular case this would give calibration factors of approximately
6.25 and 6.28 lbf ft/division. The correctness of such a decision was confirmed in
Case IV when the calibration curve of similar shape was treated in the same way.
The results are discussed below.

Case IV
During the gauging and calibration o± these shaft inserts, the opportunity was
taken to determine the effect of the proximity of the flange upon the gauge output.
These particular shaft inserts were 8 inches between flanges and five strain gauge
bridges were bonded to one shaft, the gauge positioning being at equal intervals
along the shaft The shaft was loaded torsionally and. each bridge output recorded.
Only the bridge nearest to the flange gave a significantly different output of the
order of 2%. The others gave readings which were within the tolerance of the
gauge factors for this particular set of gaiges.
The twelve shafts were calibrated in the rig described for Case III and then
three of the shafts were used during engine trials on a test bed.. The calibration
results for these three shafts and the comparison of brake and torsionmeter
readings are shown in Fig.12 and tabulated in Fig.13.
The shaft static calibration shows the now expected shape for this type of
calibration, and it was decided to extrapolate the best fit straight line obtained
from the higher portion of the calibration, to derive the lower portion. This
calibration line was used to derive the torque values for the brake/torsionmeter
tests. In this comparison only four torque values were more than 1% in disagreement
with the brake results.
The biggest difficulty with this type of comparison is the difference in
response of the two systems. The brake recording dial is deliberately damped to
give a stea&y reading and this damping produces a significant delay in recording
- 12 -

changes in torque. Since the strain indicator is read at the same time as the
brake dial, errors must occur, This is the probable reason for the apparent
scatter of results even at apparently constant brake torque conditions.

Conclusions
Methods of measuring ship shaft torque have always been a controversal
subject, each method having its own champions. However a number of systems
produce equally accurate results when installed by skilled technicians. The
general accuracy of all the systems is in the order of ±2% of maximum torque and
under favourable conditions ±1% is possible. This is probably the limit if the
ship shaft is used as the spring element. Greater accuracies could be achieved by
special calibrated shaft inserts, but the production of such units for large
torque values would be costly.
The advantages of the strain gauge system lie in other directions and the
more important can be listed as follows:-

Shaft size
Many torsionmeter systems are not suitable for attachment to shafts of small
diameter. The strain gauge does not suffer from this limitation. Slip rings of
the commercial type can be made or adapted to suit a specific shaft size, but the
temporary ring constiuction is limited to a minimum diameter of approximately
3 inches unless pre-formed during manufacture.

Installation time
At first sight it would appear that a "clamp on" torsionnieter would be quicker
to install than a strain gauge system. However in the author's experience this is

not the case. On one occasion the author and one assistant fitted two complete
strain gauge torsionmeter systems in less time than it took to install a "clamp on"
torsionmeter fitted to the same shaft.
If slow drying glues are used the installation time will take 2k hours
although the actual time "on site" is only about 6-8 hours. üth quick drying
glues or impact adhesives the total "on site" is, obviously, 6-8 hours, These

times assume that all the necessary facilities are available.

Skilled labour
All torsionmeter systems require skilled labour to ensure a proper

installation. Strain gauges are used on an amazing variety of "field" experiments


and because of this there is probably a greater labour pooi capable of fitting a
strain gauge torsionmeter than all other torsionmeter systems put together.
- 13 -

+. Cost
The price of a torsionmeter installation will depend upon the sophistication
of the system used. However, in its simplest form, the strain gauge torsionmeter
is cheaper than any other system.
Such a system would employ D.C. voltage bridge excitation with the output
measured using a D.C. microvolt meter calibrated in iv/v or direct strain. With
temporary slip rings such a system would have a total cost of less than 1OO plus
installation costs. Such a system was adopted by the White Fish Authority and
described by them in Reference 5.

Rotational speed
The gauges, having negligible mass, are neither affected by high rotational
speed nor influence the shaft. The shaft speed is limited only by the behaviour
of the slip ring assembly. Since the ring mass is small and balanced about the
shaft axis, the ring speed is limited only by the rubbing speed between ring and.
brush. Modern techniques are such that "noise free" running is possible at speeds
far in excess of any likely ship shaft speed.

Reliability
It is likely that there have been more experiments abandoned due to strain
gauge system failures than any other type of electrical strain or displacement
measuring device. Because of this the strain gauge has received much bad
publicity. This is unfortunate since the majority of failures have occurred when
no other system could have been used to perform the required task. These failures
are usually caused by installation damage, mainly wiring, and it is this type of
failure which has caused much of the criticism of strain gauge systems. In fact
the author cannot remember an occasion where failure could. be directly attributed
to gauge failure unless due to damage from some external source.
A strain gauge torsionmeter suitably protected rtfl give a high degree of
reliability, with the additional advantage that a second gauge bridge can be
bonded to the shaft to be available in the event of damage. This extra strain
gauge bridge takes little extra time to install and one bridge can be used as a
measuring check on the other to confirm satisfactory installation of both systems.
The strain gauge system, like all others, has a number of disadvantages, and
for satisfactory operation the effects of these disadvantages must be reduced to a
minimum:-
(a) Damage during installation
A strain gauge is a fragile device and is easily damaged during the
installation period. Every care must be taken to ensure adequate protection at

all times.
- 1k -

Earth resistance
The strain gauge bridge will cease to measure accurately or even have a
linear calibration if the bridge earth resistance is allowed to fall below 100
megohms. This can lead to difficulty during installation unless adequate
precautions are taken.
Brush gear damage
The recommended use of cantilever springs for the brush gear requires that
adequate protection be given to them since they are fragile.
Machining accuracy of the shaft
In the case of shafts which are oval or for hollow shafts where the bore is
not concentric, errors occur in trying to establish the values of torque from the
strain data. Under these circmtances calibration is the only suitable solution
in obtaining accurate results.
Cable resistance
If long runs of cable are used due allowance must be made for the voltage
drop along the bridge input supplies. This was necessary in Case IV where the
recording station was situated on the bridge of the ship. This is done by
comparing the gauge bridge resistance to cable resistance and correcting the strain
reading proportionately.

Acknowledgement
The work described above has been carried out at the National Physical
Laboratory.
The author wishes to acknowledge the co-operation and assistance of the
following organisations -

Bristol Siddeley Engines Ltd.


British Ship Research Asfociation.
D. Napier and Sons Ltd.
National Institute of Oceanography.
Vosper Ltd.

The author is also indebted for the help of staff of Ship Division, NPL; in
particular, to members of the Equipment Group.
- 15 -

References

(i) COOK, R. 1951 Marine torsiozimeters and thrust meters. Trans. I. Mar. E.,

Vol.63, p,ll5.

MORRISON, T. 1966 Recent developments in the measurement of propeller


shaft torque and thrust. Triu. I. Mar. L, Vol.78, p.193.

COIJTTS, J.A. 1967 Manufacturers methods of calibrating bonded


resistance strain gauges. Journal of the British Society of Strain
Measurement, Vol.3, No.2.

(1,) SIL1TERL]AF, A.., MARWOOD, W.J., BOYLE, H.B. 196k Some ship and model
measurements of unsteady propeller forces. Trans. R.I.N.A., Vol.106,
p.klS.

(5) BENNETT, R. and HATFIELD, M. 1966 Development of strain gauge techniques


for the measurement of propeller shaft torque in distant water trawlers.
The Journal of Strain Analysis, Vol.1, No.2.
FIG. I.
FIG. 2.. CONVERGING LINES AT 1° SEPARATION

FIG. 3.
CASE NUMBER 1 2 3 4
OCEANOGRAPHIC HIGH SPEED HIGH SPEED HIGH SPEED
TYPE OF SHIP RESEARCH ROUND BILGE liOft. MOTOR PLANING
VESSEL CRAFT BOAT CRAFT
SHAFT DIA. 9 InS 51/4 InS 49x 40(HOL) 51/4X 31/4(HOL)
Z MATERIAL STEEL. STEEL STEEL STEEL
0
MAX. R.P.M. 190 42.0 1,000 1,700
4
o MODULUS OF
RIGIDITY 11B6x106 UNKNOWN UNKNOWN UNKNOWN
o (lbf/1n2)
w MAX. LINEAR
a.
STRAIN 220/..Le 530ue 570,u.e 25o,e
-AU GE
I-
LI. MAX. SHEAR
4 STRESS 5,300 12,700 13,700 6,000
U)
(Ibf/Ifl2
'
2
N OF SHAFTS 1

PAPER
2
PAPER
= 3

GAUGE EPDXY EPDX.Y


IRIDIUM KARMA FOIL FOIL
_j TYPE
PLATINUM WIRE WIRE
4
I- GAUGE 6 2 2 a
a FACTOR
N°' MEASG. 2 1.- PERSHAFT a- PER SHAFT 2-PER SHAFT
z BRIDGES
0
4
BRIDGE
NER.GISATI0N
1,000 C/s AT
5-VOLTS
1,000 c/s AT
5-VOLTS
2,500 c/S AT
S-VOLTS
* 1,000 AND
2,500 cls
AT 5-VOLTS
-J
PHILLIPS
GAUGE ARALDITE ARALDITE ARALDITE
STRAIN GAUE 753 753
GLUE
CEMENT
TWIN PACK
z
- SLIP RING COMMERCIAL
SPLIT
COMMERCIAL
SPLIT
COMMERCIAL
SPLIT E4 CONSTD
COMMERCIAL
SPLIT
TYPE
METHOD
OF
A.E.1. SIMPLE CALIBRATION
CALIBRATION TORSIONMETER CALIBRATI3N (FIG.9.) PLUS
OR CA LI B RATION
FITTED BY RIG CONSTD CHECK ON
COMPARI SON
B.S.R.A. 'ON SITE' RIG, (FIa9.) TEST BED
OF
RE SULTS
OUTPUT SPECIAL SPECIAL
SPECIAL MEASD. BY NONE SHAFT SHAFT
0 M ME NIS INSERTS INSERTS
NULL BALANCE

ALL FIGURES QUOTED ARE NOMINAL VALUES


* DIPFERENY INSTRUMENTS USED DURING INITIAL CAL.
AND TEST BED TRIALS

FIG. 4. SUMMARY OF CASE HISTORIES


0
8,500 X STRAIN GAUGE BRIDGE. 1.
STRAIN AUE BRIDGE. 2.
o A.E.1. TORSIONMETER
8,000
('I

7,500
Ui
±10

7,000
Lii

0-i
0>
6,500

6,000
0 10 ao 30
SHAFT REVOLUTIONS/ SEC (n)

FIG. 6. COMPARISON OF STRAIN GAUGES AND A.E.1. TORSIONMETER

CASE HISTORY I
z
0
lii
1O5

uI<
'czW
a:
I04
tow
ow
103
I-U,
jg
wo
0

0
I-
101

100
30 40 50 60 70 80 90 100 110
TORSIONMETER READING (DIVISIONS)

FIG 7. CORRECTION FACTOR USED FOR A.E.1. TORSIONMETER

CASE HISTORY I
X SHAFT. 1.
-I- SHAFT. a.
1700

1650

1.00

1550

1500
200 400 600 800 1,000 1,200 1,400 1,600 1,800
READING (DivisioNs)

FIG, 8. CALIBRATION OF SHAFTS

CASE HISTORY II
LIFTING JACK.

- HYDROSTATIC
CAPSULE.

LOADING GUIDES.

////////////////////////////////// ///
FIG 9A SIDE ELEVATION

FIG 9.8. END ELEVATION


670
0
O SHAFT () CAL.1.
z El SI-IAFT () CAL.a. ELAPSED TIME BETWEEN
2 660 x CALIBRATION .1 E CALIBRATION.2.
X SHAFT ® CAL.1. .

> + SHAFT ® CAL. a. APPROXIMATLY 2 MONTHS


a
4-
+
9- 650
"S
El

640 0
0
I-
x
0 + SHAFT.a.
LL
630

9 0 1g
620 x 0
SHAFT.1.

2.00 400 600 800 1,000 1,200 1,400 1,600 1,800 2.,000
BRIDGE OUTPUT (DIVISIONS)

FIG. 10. SHAFT CALIBRATIONS

CASE HISTORY III


RESULTS ARE FOR THE TOTAL
POWER FROM BOTH ENGINES

1,000 2,000 3,000 4003 5,000 6,000


N.P.L. POWER (HP)

FIG. 11. COMPARISON OF tLP.L. AND ENGINE MANUFACTURERS POWER MEASUREMENTS

CASE HISTORY III


1250
+ BRIDGE SHAFT a

1240
+
x
o
- BRIDGE A. SHAFT 3
BRIDGE

BRIDGE
B.

B.
SHAFT 3
SHAFT 12

1230
0

ta20 0
w
0 D
0
I-
1210
tL
0
w
z
0 --0- J
0
1200 a-

U-
-J 0
4-)

iiqo 0'
'-4

.1-I

100 200 300 400 500 600 700 600 900 10O
BRIDGE OUTPUT (DIVISIONS)

FIG. ia. CALIBRATION OF SHAFTS

CASE HISTORY
(THESE SHAFTS USED FOR ENGINE TEST BED TRIAL)
FIG. 13

RESULTS OF ENGINE TEST BED TRIAL


ON THREE SHAFTS CASE IV

R.P.M.
(R TORQUE
TORQUE TORQUE
BRAKE TOR W
x ioo)%
lbf ft lbf ft
SHAFT 3 BRIDGE A

695 1826 1846 +1.07


715 362'-F 3606 -0.1+8
752 5+9O 51+1+3 -0.86
752 51+65 51+30 -0.66
751 5610 5581+ -o.k6
900 721+3 719k -0.62

SHAFT 3 BRIDGE B

1050 8226 8205 -0.25


1050 8262 8225 -0.1+1+
1202 9221 9132 -0.98
1350 10355 101+39 +0.81
1350 10211 10288 0.75
1520 9996 ioiok +1.02
171+3 12323 12371 +0.39
171+3 121+51+ 12502 +0.38
171+3 12570 121+37 0.51+
171+3 12311 12371 0.1+8
17+3 12358 121+10 0.1+2
171+3 12299 121+10 +0.89
171+3 12335 12397 +0.50
171+3 12323 12381+ +0.1+9
171+3 12323 12401+ +0.65
1525 12335 12239 -0.78
1525 12323 12338 +0.12

SHAFT 12 BRIDGE B

1352 10120 9993 -1.20


1525 10022 10141+ +1.21
1525 1191+3 12010 +0.51+
1683 12099 12187 +0.72
171+3 12321 12371 +0.59

SHAFT 2 BRIDGE A

1350 io86 10656 +0.65


1525 10112 10190 +0.76
1525 12001 12111 +0.91
1683 12001 12111 +0.91
171+3 12298 1231+5 +0.38
z

THE TOR.QUE ARM AO, WHEN LOADED BY A


ZERO FORCE.' FORCE F ROTATES TO A NEW POSITION A10
POSITION DUE TO TWISTING OF THE SHAFT. SINCE THE
JACKING AXIS IS FIXED, THE EFFECTIVE
TORQUE ARM IS
ie. APPARENT TORQUE =Fx
ACTUAL TORQUE =Fx e1 F x C COS O(

(SMALL ERROR DUE TO TORQUE ARM FLEXING


IS NOT SHOWN HERE
0
A
SHAFT
AXIS

IG.14. EFFECTIVE CHANGE OF LENGTH OF


SHAFT CALIBRATION RIG TORQUE ARM.

You might also like