Você está na página 1de 5

Zero-shift Values of Automatic and lnexpensive

Strain-gage lnstrumentation Systems


by P. Cappa, K.G. McConnell and L.W. Zachary
ABSTRACT -Computer-controlled data-acquisition systems
are being used extensively for gathering strain-gage data. This
paper explores the relative merits of using modern solid-
state digitai multi-meters (DDM) to measure the strain-gage
resistance directly rather than using a conventional Wheat-
stone bridge. Both a direct-resistance measurement scheme
and a reversed current scheme are compared aver long
measurements terms of 6 and 12 days using a 6V2-digit
multimeter. The results show that the reversed current method
is superior in maintaining the zero-gage resistance reading at
the cast of using severa! meters. Possibly the direct-resistance
method can be improved so that the operation and equipment
can be simplified.
lntroduction
Using computer-controlled data-acquisition systems in
both laboratoryl-l and field-testing situations is in-
creasing. Some of the automated, multichannel, sequential
strain measuring systems7-l' use intermediate circuitry
between the multiplexer and the A/D converter. The
circuitry is capable of minimizing the errors resulting from
the random variations of the voltage drops caused by
switch and leadwire resistance. The variations of these
parasitic resistances cause nonreproducible readings,
unstable zero readings, and loss of caiibration. These
systems aiso have the capability of suppressing thermo-
couple voltages generated by the switching units.
One method of reading a large number of strain-gage
data in a successive manner is to use a switching unit so
that one measurement instrument can accommodate
several gages. The following automated arrangements
have been described and tested in literature: a Wheatstone
P. Cappa (SEM Member) is Assistant Professar, University oj Rome,
"La Sapienza," Department of Mechanics and Aeronautics, 00184 Rome,
ltaly. K.G. McConnell and L. W. Zachary (SEM Members) are Projes-
sors, Jowa State University, Department oj Engineering Science and
Mechanics, Ames, lA 500ll.
Paper was presented at the 1990 SEM Spring Conjerence on Experimental
Mechanics held in Albuquerque, NM on fune 4-6.
Originai manuscript submitted: Ju/y 2, 1990. Fina/ manuscript received:
December 5, 1990.
88 March 1991
bridge out of balance
12
and a direct measurement of the
gage resistance with two- and four-wire connections.'
3
-
16
The present paper will examine the limits of stability of
two sequential automated systems that use a four-wire-
connection wiring scheme. The first system directly
evaluates the resistance variations by using a single digitai
ohmmeter, while the second system uses a reverse-current
generator and several digitai meters. In the second system,
the current (both positive and negative values) and the
corresponding voltages are measured simultaneously to
determine the strain-gage resistance readings. The value of
zero shift is calculated from these readings.
The devices are interfaced by means of the IEEE-488
bus to a computer that controls the measuring instruments
used in the test sequence and stores and analyzes the data.
The tests were organized so that 100 measurements sets
were taken every 20 minutes. The mean of the resistance
vaiues obtained from the first hundred resistance readings
was taken as the reference resistance value and then the
zero-drift values were calculated on the basis of this
reference value. A similar approach was used for the
reverse-current method.
Direct-resistance Method
The experimental test setup for the direct-resistance
measurement method is shown in Fig. l.
The strain gage under test, indicated by R, in Fig. l, is
connected to the Keithley Model 196 digitai multimeter
(DMM) by four leadwires as shown. The 'Hi Lo' connec-
tions drive a constant current of approximately l. 7 mA
through the gage while the sense connections measure the
voltage drop that develops across resistance R
8
However,
this measurement method does not cancel factors such as
thermoelectric and electrochemical effects that can
influence measurement accuracy. The strain gage, whose
resistance is nominally 120 O, was connected by wire leads
approximately 1-m long with a diameter of 0.4 mm.
The Modell96 DMM is a 6\/z digit unit with a resistance
resolution of 100 ~ t O when set in the 300 range to
measure a nominai resistance of 120 n. This resolution is
equivaient to 0.42 f!m/m. This unit's offset-compensated
ohms feature can compensate up to 30 m V of thermal
emf's at the test resistance.
The !oca! temperature at the strain-gage location is
measured simultaneously by using a chromel-alumel
thermocouple, TC. The thermocouple is connected to an
ice point, IP, and a Keithley Mode! 195 DMM as shown.
The Mode! 195 is a 5 Y2 digit uni t with a resolution of
1.0 f1- V an d an input impedance greater than 1.0 G!]. The
two digitai multimeters are connected to a computer
through the IEEE-488 data bus. Data are stored on a
floppy disk and on a printed hard copy.
The data raws relative to the strain-gage resistance and
the test-area temperature were collected every 20 minutes.
Measurements, as is well known, are mostly random in
nature, so one way to overcome this randomness is to use
the built-in filter of the DMM. Thus, these tests were
conducted in two different ways: with the filter 'on' and
with the filter 'off'. When the filter is 'on' the data of the
strain-gage resistance,

presented to the computer are


the weighted average of a number of readings, R
1
,J, K. In
this case K and J were equa! to 100. Then, 100 weighted
averages are read by the computer. From these data mean
values, R
1
, an d twice standard deviation, 2 S D R
1
, are
computed. The test-area temperature, as outlined in Fig.
2, was measured only o ne time, T1 ,h every 100 resistance
readings. From these data mean values, T
1
, and twice
standard deviation, 2 S D are determined. One single
mean resistance and a single mean temperature reading
are recorded for each time period along with twice the
standard deviation of these averaged readings for each
variable. It is well known that over 95 percent of the
readings should li e within 2 standard deviations. Sin ce
it is hypothesized that the measurements are subjected to
many small sources of random error and negligible
systematic errors, then the Gaussian distribution should
describe the data distribution quite well. The standard
deviation is calculated in accordance with the more
conservative, i.e., larger, definition.
The statistica! measurement scheme, see Fig. 2, shows
that the first reading is used as the reference reading. The
zero-shift value for the nth reading is then calculated
from the relationship
H i
r------kl Sense
Keithley
Mode l
196
DMM
Lo w
Keithley
Model195
DMM
D
Computer
Fig. 1-Schematic of direct-resistance
measurement method
(l)
w h ere a gage factor, F, of two (2) is assumed: R
1
is the
first averaged reading and R. is the nth average resistance.
For the filter 'off' case only one reading is used (K =
l) in piace of the weighted average

of 100 readings.
The obtained data, obviously, will show more scatter
because of the loss of the filter effect from the averaging
scheme.
Reverse-current Method
The reverse-current method has the advantage of
cancelling the effects of any d-c offset voltage that occurs
because of wire-connection thermocouple effects. This
measurement method uses the equipment shown in Fig. 3.
In this case, a Keithley Mode! 220 reversible current
source is used to drive current through the leadwires and
gage resistance. In order to check on the accuracy of the
current source, one Keithley Mode! 196 DMM is connected
to measure the current while a second Keithley Mode!
196 DMM is connected across the strain gage to measure
the voltage drop over R .
The ambient temperature near the gage is measured by
using the same thermocouple and the Keithley Mode!
195 DMM as used previously.
Ali instruments are controlled over the IEEE-488 data
bus by a computer. The data are stored on a floppy disk
as a permanent record.
The reversed-current method is based on the following
equations. The voltage-current relationships for the strain-
gage resistance R. are given by
VP = I P R. + A V. (2)
for the positive current. For the negative current
VN= INR.+AV_ (3)
where VP is the positive-voltage reading, lP is the positive-
current reading, A v. is the positive-current error voltage,
FILTER "DN"
REFBRENCE VALUE
Rl, 1,1
Rl,l
Rl,l,lOO
Rl ::_ 2 SD R
1
Rl,lOO, l
Rl,lOO
Rl, 100,100
n TH DATA LOOP
Rll, 1,1
l
Rn,l, 100
R n..:!:. 2 SD Rn
R
n, 100,1
Rn,lOO
R
n, 100,100
Fig. 2-Scheme for analyzing direction-resistance data
Experimental Mechanics 89
V N is the negative-voltage reading, l N is the negative-
current reading, L1 v_ is the negative-current error voltage.
Subtracting eq (3) from eq (2) gives
R = ( VP- VN) + flR,
(lP- IN)
(4)
where t.R' is the error due to (fl V,- L1 V_) being non-
zero. However, t.R' should be nearly zero if thermo-
electric effects remain the same for both positive and
negative currents. Hence, the reversed current method is
capable of significantly reducing this source of error. Note
that VN and IN are negative numbers so that R, = V l l.
TC
Keithley
Mode l 196 DMM
Voltage
Keithley
Mode! 196 DMM
Current
lP ~
IEEE-488
Data Bus
Keithley
Model22.0
Current
Source
D
Computer
Fig. 3-Schematic of reversed-current
measurement method
REfERENCE VALUE
IPl, l
VPl, l
IN l, l Rl, l
VNl,l
,
.. 2 SD R
1
IPl,SO
VPl ,50
IN1,50 Rl ,50
VN1,50
n TH DATA LOOP
IP
n, l
VP
n,l
INn,l
R
n, l
VNn,l
R + 2 SD R
n
IPn, 50
VPn,SO
INn,SO Rn,SO
VN
n,SO
FIGURE 4
Fig. 4-Scheme for analyzing reversed-current data
90 March 1991
The statistica! scheme for the reversible current measure-
ments is shown in Fig. 4. In this case, a complete set of
data consists of the positive and negative voltages and
currents and the temperature. These voltage and current
readings are used to calculate the effective resistance
R.,, through R., so by using eq (4) with t.R' assumed t o
be zero. These 50 resistance values are used to calculate
the mean readings R. and the corresponding twice standard
deviation. The apparent strain is then calculated using
eq (l).
Test Results
Seven different sets of test results are presented here.
The test conditions are summarized in Table l. The first
three tests were the direct-resistance measurement type,
while the last four tests were the reversed-current type.
During the direct-resistance methods tests, the Keithley
Mode! 196 DMM supplied a current of approximately
1.7 mA to the strain gage. This DMM has an auto-zeroing
feature that zeros the meter with no current flowing, a
feature that is important in reducing data scatter.
20
.........
E 1o
"-..
E
"' l
o
o
~ 1 0
_c
m
l
e -20
(j)
N
48 72 96 120 144
Time (hours)
Fig. 5-Zero-shift data during test #1 (direct-resistance
method with filter 'on')
20
E' 10
"-..
E
"' o
l
o
...... -10
:-t:
_c
m
e -20
(j)
N
48 72 96
Time (hours)
Fig. 6-Zero-shift data during test #2 (direct-resistance
method with filter 'o n')
168
144
Tests one and two are duplicates and each lasted for
approximately 140 hours. For these two tests, the internai
filter was 'on'. Figures 5 and 6 show that the zero shift
tends to decrease with time during test o ne (slope -0.137
per hour with a high correlation coefficient of
-0.903), while in test two the results have essentially
zero slope (slope -0.03 11m/m per hour with a low
correlation coefficient of -0.288). The mean reading for
test one is -2.62 11m/m with a standard deviation of
6.50 11m/m, while for test two the mean reading is
5.45 11m/m with a standard deviation of 4.10
11
m/m. The
mean reading is, obviously, highly influenced by the value
obtained for the first reading. However, the standard
deviation from the mean should be nearly the same
regardless of the value of the first reading. In this case,
the standard deviation is greater for data set one than for
data set two.
The third data set was obtained by using the direct-
resistance measurement method with both the filter
feature an d the auto-zero feature turned 'off'. The results
of this test are presented in Fig. 7. The instrument
showed signs of significant drift in this case because the
constant-strain condition showed nearly 30- 11m/m drift in
66.7 hours. There is no apparent correlation of data from
tests one through three with temperature.
TABLE 1-TEST EXPERIMENTAL PROCEDURE
Zero Shift
Test Test Type No. Data Hours Mean Std.Dev.
No. Points of Test [.umlm] [.umlm]
1 Dir.Res.Filter 'on' 448 149 -2.26
2 Dir.Res.Filter 'on' 414 138 5.45
3 Dir.Res.Filter 'off' 200 66.7 -10.50
4 Rev.Curr. (20 mA warm) 250 83.3 2.15
5 Rev.Curr. (20 mA cold) 46 15.3 9.97
10.73
6 Rev.Curr. (5 mA cold) 150 50.0 -0.39
7 Rev.Curr. (10 mA cold) 146 48.6 2.87
originai data has ramp increase during first 1.7 hours
tMean va/ue and standard deviation alter 2.0 hour warm up
20
r--.
E 10
'-..
E
<D
l
o
o

..c
m
l
2-20
Q)
N
24 48
Time (hours)
6.50
4.10
4.30
3.80
2.51.
1.16t
2.89
2.37
72
Fig. 7-Zero-shift data during test #3 (direct-resistance
method with filter 'off')
The zero-shift results for the reversed-current measure-
ment method are shown in Figs. 9-11 for tests four through
seven. These results were obtained by using reversed
currents of 20 mA (gage warm and cold), 5 mA (gage
cold) and 10 mA (gage cold). The heating effects were
obvious during the first 1.5 hours for test five (see Fig. 10)
where the zero-shift strain increased about 10 This
gage-heating phenomenon is also evident in the mean
values and standard deviations shown in Table l. Tests
six and seven were conducted at 5 and 10-mA currents
with the gage cold in order to detect a workable lower
limit. The graphs show that the heating for both 5-mA
and 10-mA currents was not significant compared to the
20-mA results. Examining the mean and the standard
deviation data in Table l leads to the same observation.
Thus, one concludes that the heating effect is negligible
for the 5-mA case and only marginai for the 10-mA case.
Although this conclusion is limited to experimental
situations similar to the one used here, it is apparent that
20
r--.
E 10
'-..
E
<D
l
o
o

..c
m
l
2-20
Q)
N
24 48
Time (hours)
Fig. 8-Zero-shift data during test #4 (reversed-current
method with 20 mA and strain-gage warm)
20
r--.
E 10
'-..
E
"'
o
l
o

..c
m
l
2-20
Q)
N
Time (hours)
Fig. 9-Zero-shift data during test #5 (reversed-current
method with 20 mA and stain-gage cold)
24
Experimental Mechanics 91
the lowest current causes the least amount of heating
effects.
One of the effects expected is the correlation between
stran readngs and temperature. The relationship between
zero-shift strain and temperature for the different tests
shows a nearly random behavior with only a slghtly
negative slope. Graphs of this kind indicate that the errors
are random relative to temperature.
Conclusions
The values of the zero shift obtained by using direct-
resistance measurements confirm the results available in
the literature." '
5

16
The drift remained within 15 ~ t m m
for the filter and auto zero both 'on' (test l and 2 in
Figs. 5 and 6) for tests that lasted approximately 140
hours with a temperature variation ranging from 21 o C
to 24C.
By using the reverse-current method, nominai zero-
shift values were obtained when the current level did not
20
,___
E 10
"'-..
E
<O
l
o
o
;:::: -10
_c
m
l
e -2o
Q)
N
24 48
Time (hours)
Fig. 10-Zero-shift data during test #6 (reversed-current
method with 5 mA and strain-gage cold)
20
r--..
E 10
"'-..
E
<O
l
o
o
;; -10
_c
m
l
e-20
Q)
N
24 48
Time (hours)
Fig. 11-Zero-shift data during test #7 (reversed-current
method with 10 mA and strain-gage cold)
92 March 1991
72
72
cause significant heating effects in the strain gage (see
Figs. 8-11). The threshold current level appears to be
between 5 and 10 mA for the gage-installation tested. The
effect of current level on the zero shift is demonstrated.
The standard deviation of the zero-shift data was smaller
for the reversed-current method than for the direct-
resistance measurement method. The temperature range
was found to be the same for the reversed-current test
conditions.
The results from these two systems suggest a low-cost,
general-purpose system that is capable of providing small
zero-shift values in long-term stran-measurement situa-
tions: the reversble-current system. Unfortunately, the
reversed-current method requires more measurement
equipment than the direct-measurement method. The
possibility that the direct-measurement method may be
modified to achieve comparable or better measurement
accuracy is being investigated. Such a development would
be an improvement in simplcity of operation and is
being investigated.
Acknowledgments
The authors appreciate the support of the Department
of Engineering Science and Mechanics and the Engineer-
ing Research lnstitute that provided the necessary facilities
to conduci this research. In addition P. Cappa wishes to
acknowledge the financial support provided by the Italian
National Research Council (Grant 203.07.16).
References
l. Jedju, T.M., "A Data Acquisition and Contro/ System Based on
the New Commodore Pet Microcomputer," Ref. Sci. Instru., 50 (9),
1077-1079 (1979).
2. Estler, R.C., "Data Acquisition and Contro/ System Based upon
the Rockwe/1 AIM-65 Microcomputer," Rev. Sci. Instr., 51 (10), 1428-
1430 (1980).
3. McConnell, K. and Abdelhamid, H., "Un/eashing a Microcomputer
in the Laboratory," EXPERIMENTAL TECHNIQUES, 7, 22-24 (Jan.), 38-40
(Feb.), 28-30 (March 1983).
4. Lu, W. Y., "A Personal Computer Base Testing System," Proc. V
lnt. Cong. on Exp. Mech., 245-249 (1984).
5. Lu, W. Y., "A Personal Computer Base Testing," Proc. 1985 SEM
Fa/l Conf. on Exp. Mech., Greneleje, 141-145 (1985).
6. Teajord, W.J., "Computer Fie/d Data Acquisition," Proc. SESA
Spring Meeting, 26-27 (1976).
7. Kreuzer, M., "Comparing the Efject of Lead and Switch
Resistances on Vo/tage- and Current-jed Strain-Gage Circuits," Rep.
App/. Measurements, l, 13-18 (1985).
8. Howland, H.J., "An Integrated Software/Hardware Approach to
Experimental Stress Analysis," Proc. VII Int. Conf. on Exp. Stress
Analysis, 263-270 (1986).
9. Kreuzer, M., "How to Avoid Errors Caused by Heat Efjects in
Strain Gage Measurements when Using Scanning Units," Proc. VII Int.
Conj. on Exp. Stress Analysis, 447-454 (1986).
10. Versnel, W.J., "Compensation oj Lead-wire Ejfects with Resistive
Strain Gages in Mu/ti-channel Strain Gauge Instrumentation, " Proc. VII
Int. Conf. on Exp. Stress Analysis, 455-464 (1986).
11. Boswe/1, R. S. and Payne, M.L., "Deve/opment and App/ication oj
Special Purpose Data Acquisition System jor Strain Gage Testing of
Threaded Tubular Connections Used in Oi/field," Proc. 1987 SEM Spring
Conf. on Exp. Mech., 158-163 (1987).
12. Williams, M. and McFetrdge, G., "Unbalanced-bridge Computa-
tiona/ Techniques and Accuracy jor Automa/ed Multichannel Stran
Measuring Systems," EXPERIMENTAL TECHNIQUES, 7, 32-37 (1983).
13. Nelson, E.J., Skorra, C.F. and Howard, J.L., "Measurng Stran
Gages Direct/y Without Signa/ Conditioning," EXPERIMENTAL TECH-
NIQUES, 7, 26-28 (Sept. 1983).
14. Hillen, R.E. and De Witt, R.J., "Measuring the Coejjicient oj
Thermal Expansion oj Composi/es Using the Strain Gage Method," 1987
SEM Spring Conf. on Exp. Mech., 158-163 (1987).
15. Cappa, P., "Limits of lnaccuracy oj a Microcomputer Contro/led
Scanning System for Strain Gage Data Readings," VI Int. Cong. on Exp.
Mech., 469-473 (1988).
16. Cappa, P., "Limits of Inaccuracy oj Automatic Direct Systemsjor
Strain Gage Data Readings Based on a Low Cost Switch Contro/ Unit, "
EXPERIMENTAL TECHNIQUES, 13-15 (Sept. 1989).

Você também pode gostar