3 Eurolab Dahlberg

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

EUROLAB International Workshop: Investigation and Verification of Materials Testing Machines 21

Materials Testing Machines


Investigation of error sources and
determination of measurement uncertainty
G. Dahlberg; MTS Systems Corporation, Eden Prairie, USA

In an effort to meet National, Inter- uncertainty. The task then is to identify as


national, and Commercial Accreditation many sources of error as possible, and take
requirements, users of Materials Testing measures to reduce or at least quantify, the
Machines must establish adequate methods resultant measurement uncertainty of data
of determining calibration and test measure- and or results produced by the material test-
ment uncertainties. Determining measure- ing machine. Under testing a material or
ment uncertainty for testing machines and component may lead to safety, warranty, and
test data can be a very complex process. liability problems due to premature failure or
Many factors affect the uncertainty of test damage. This is of particular concern for the
data produced by testing machines. transportation and medical industries. Under
testing conditions exist when the testing
This paper examines measurement machine end level forces are not achieved
uncertainty contributors associated with the and or the test speed, frequency, or cycle
calibration and use of modern materials test- count is less than required to meet testing
ing machines. A list of uncertainty contribu- criteria. Over testing a material or compo-
tors is provided including examples of error nent may lead to waste of time and material
source values and their assumed statistical for design, fabrication, and test. Over testing
distribution types. Methods of minimizing is expensive and can potentially reduce
certain error sources are described. Also competitive advantage. An over testing con-
included, is a short section discussing mea- dition typically occurs when forces exceed
surement uncertainty related to ISO calibra- the testing criteria. Under testing and over
tion practices ISO 376, Metallic materials – testing conditions can occur when accelera-
Calibration of force-proving instruments tion induced forces are present, during tran-
used for the verification of uniaxial testing sient cyclic overshoot of a start up wave-
machines and ISO 7500-1, Metallic materi- form, when a testing machine is poorly
als – Verification of static uniaxial testing controlled, or when the system is misadjust-
machines – Part 1: Tension/compression ed or out of calibration status. So aside from
testing machines – Verification and calibra- regulatory accreditation pressures, currently
tion of the force-measuring system being imposed on all areas of test and mea-
surement, there are other excellent reasons
This paper discusses the effect of calibra- for examining measurement uncertainty re-
tion uncertainties on test data produced by lated to the operation of material testing
materials testing machines. Examples of how machines.
operators, materials, testing methods, and
testing machine limitations, contribute to the One common misconception and impor-
measurement uncertainty of test data. tant point of interest, is that Testing Machine
Manufacturers can provide total or combined
One of two conditions will exist when measurement uncertainty values for material
testing materials or components. The materi- testing machines. This is simply not true and
al or component will either be under tested it would be unwise for manufacturers to pro-
or over tested. There is simply no way to per- vide these values. Testing Machine manu-
form a materials test that contains zero error facturers design systems and software to
or has a zero value for the measurement perform specific tests under specified condi-
22 EUROLAB International Workshop: Investigation and Verification of Materials Testing Machines

tions. Many specifications are published as Major sources of measurement uncertain-


optimal operating specifications. Few if any ty can be grouped into the following cate-
systems are operated under optimal condi- gories.
tions. Sources of measurement uncertainty
should be assessed experimentally for each 1) Uncertainty due to the calibration equip-
type of material, system configuration, and ment and calibration processes
testing protocol to be performed. 2) Uncertainty of the Testing Machine as
calibrated
This paper concentrates on the indication 3) Uncertainty of the Testing Machine
and application of force when performing during use
tests using material testing machines. The 4) Uncertainty of the Test Results
author acknowledges that material testing
machines are used to apply and measure It must be understood that the uncertain-
additional physical metrology parameters ty values presented in these examples are
that may include but are not limited to dis- representative of a particular uncertainty
placement, torsional force, angular displace- analysis and do not represent all material
ment, pressure, and strain. An uncertainty testing machines of any specific type and or
analysis pertaining to any metrologically sig- configuration. It should also be recognized
nificant parameter applied or measured, and that various interpretations related to error
reported by the testing machine should be sources and their statistical contributions
assessed. may vary depending on the method of analy-
sis applied.

Static Calibration Uncertainty

(1) ISO 10012-1: 1992(E), Calibration – The set of operations which provides information so that adequate ad-
ISO 7500-1: 1999(E)
establish, under specified conditions, the re- justments can be made if required. Perform-
(2) ANSI/NCSL Z540-1,1997. lationship between values indicated by a ing a calibration does not always require an
Calibration – The set of measuring instrument or measuring system, adjustment.
operations, which establish,
under specified conditions,
or values represented by a material measure
the relationship between or a reference material, and the correspond- Uncertainty Static Calibration:
values indicated by a mea-
ing values of a quantity realized by a refer- Usc = v 0.052 + 0.252 + 0.042 + 0.022 +
suring instrument or measu-
ring system, and the corre- ence standard.(1) (2) 0.022 = 0.26%
sponding standard or (Root Sum Squared Method – RSS)
known values derived from
the standard.
The term Calibration has often been
associated with the act of making adjust-
(3) Standard Uncertainty
ments. When in fact the calibration process
express as one standard
deviation or one sigma.

(4) Statistical Distribution.

Reference ANSI/NCSL Source Uncertainty (3) Distribution (4)


Z540-2-1997, NIS 3003 Primary Force Calibration 0.05% Class 1 (5) normal
Edition 8, May 1995
Force-Proving Device 0.25% Class 1 normal
(5) ISO 376, Metallic material
Long Term Drift of Calibration Device 0.04% rectangular
– Calibration of force-pro-
Environment 0.02% rectangular
ving instruments used for
the verification of uniaxial Process Repeatability 0.02% normal
testing machines, Table 2.
EUROLAB International Workshop: Investigation and Verification of Materials Testing Machines 23

Primary Force Proving Devices are cali- classification criteria for static force calibra-
brated in compliance with well documented tion of material testing machines.(7)
practices. Although many countries have
developed there own calibration/verification Environment is an important parameter
procedures, it is likely that most countries when calibrating testing machines. The tem-
will soon adopt ISO 376 as a common guide perature is very closely controlled in labora-
when calibrating force-proving devices. tories calibrating force-proving devices. ISO
Table 1, shows the parameters evaluated 376 requires that the laboratory temperature
during an ISO 376 calibration and associated to be maintained with in ±1 ºC.(8) The cali-
classification criteria. (6) bration is to be performed with in a tempera-
ture range of 18 ºC to 28 ºC. Adequate time
I believe that most countries will also be must also be allowed for the force-proving
adopting ISO 7500-1 as a guide when cali- device to attain a stable temperature. This
brating the static force indicating perfor- may take as long as an hour and can be
(6) ISO 376, Table 2.
mance of the testing machine. Table 2, assessed by monitoring the zero force indi- (7) ISO 7500-1, Table 2.
shows the measurement parameters and cation. (8) ISO 376, Section 6.4.3.

Characteristics of force-proving instruments Uncertaintyª


of applied
% calibration
Class of reproducibility of repeatability of interpolation of zero of reversibility force %
00 0.05 0.025 ±0.025 ±0.012 0.07 ±0.01
0.5 0.10 0.05 ±0.05 ±0.025 0.15 ±0.02
1 0.20 0.10 ±0.10 ±0.05 0.30 ±0.05
2 0.40 0.20 ±0.20 ±0.10 0.50 ±0.10
ª The uncertainty of the calibration force is obtained by combining the random and systematic errors of the calibration force.
Table 1

Characteristic values of the force-measuring system


Class of Maximum permissible value, %
Machine Relative error of Relative
range accuracy repeatability reversibility zero resolution
q b v fo a
0.5 ±0.5 ±0.5 ±0.75 ±0.05 0.25
1 ±1.0 ±1.0 ±1.5 ±0.1 0.5
2 ±2.0 ±2.0 ±3.0 ±0.2 1.0
3 ±3.0 ±3.0 ±4.5 ±0.3 1.5
Table 2
24 EUROLAB International Workshop: Investigation and Verification of Materials Testing Machines

When these devices are used to calibrate I have not included Alignment as one of
material testing machines, the environment the uncertainty contributors related to cali-
that the testing machine operates in can be bration. Calibration technicians are normally
quite different. ISO 7500-1 requires that the trained to ensure proper axial alignment
temperature during machine calibration not when doing calibrations. Most systems
(9) ISO 7500-1, Section 6.4.2. vary more than ±2 ºC.(9) The calibration of would be difficult to fixture if alignment were
the testing machine shall be carried out at in such a condition that the calibration data
an ambient temperature between 10 ºC and would be significantly affected. The force-
35 ºC. Most force proving devices have proving device is rotated and data is ac-
temperature compensation gauges or meth- quired during calibration. The force-proving
ods for calculating the reference calibration device’s sensitivity to a normally experienced
adjustment due to temperature. out of alignment condition is included in the
combined uncertainty for the device. This is
When using dead weights, the force must why fixtures and studs used with force-prov-
be compensated for the local value of gravity ing devices, should be used when force-
and air buoyancy. I have not found any proving devices are being calibrated. I did
evidence that humidity has an effect when however experience a rare situation while
performing calibrations of material testing calibrating a universal testing machine. This
systems. However, if force proving devices was a testing machine in which the system
are not adequately sealed against humidity, force transducer is physically moved when
extended time exposure to high humidity switching from Tension to Compression
environments can cause bonding of force mode or the reverse. An amount of dirt
sensing gauges to degrade. and debris present on the mating surface
between the system crosshead and force
Process Repeatability is determined ex- transducer caused an out of alignment con-
perimentally by having a number of techni- dition resulting in a 2% shift in the sensitivity
cians perform calibrations with the same of the force indication device. Therefore it is
equipment. This can be done two ways. The important to inspect these mating surfaces
technicians can use the calibration equip- prior to performing a calibration.
ment to calibrate an artifact that simulates a
system or the technicians can calibrate and Many quality control programs require
actual testing machine. If the technicians cal- that «As-Found» calibration data be obtained
ibrate a testing machine, some of the vari- prior to making any changes to a testing
ance in the data will be due to the testing machine. This is good metrology practice
machine itself. This is difficult to assess. and can provide confidence in data and tests
Automated calibration systems help reduce performed between calibration intervals. On
the variance due to process repeatability. the other hand this practice can also uncover
evidence that materials or components have
been incorrectly tested. If As-Found calibra-
tion data is required, calibration data must
be recorded prior to cleaning or changing the
condition of the machine.
EUROLAB International Workshop: Investigation and Verification of Materials Testing Machines 25

Static Testing Machine Uncertainty

Source Uncertainty Distribution


Resolution 0.20% rectangular
Uncorrected Error 0.20% rectangular

Uncertainty Static Testing Machine: ern Material Testing Machines are dynamic
Ustm = v 0.22 + 0.22 = 0.28% to a certain degree. Tests that utilize dead-
weights for simple proof testing, are the
Resolution is often defined as one half exception.
the noise, with the lowest calibration force
applied, or one digit of a digital display, Most modern material testing machines
whichever is greater.(10) provide for some type of closed loop control (10) ASTM E4-99,

Standard Practices for


of the machine. Simply described, closed
Force Verification of Testing
Uncorrected Error is the amount of error loop means that a signal from one of the sys- Machines, Sections 3.1.12,
from the static calibration of the testing tem’s calibrated transducers is fed back into 3.1.13.

machine. This amount of error could be the system’s control circuit to automatically (11) ISO 7500-1 Section 6.1,

equal to the maximum value of the error re-adjust the system to maintain desired lev- ASTM E4-99 Section 1.1.
allowable specific to the required Testing els of force and or displacement. System
Machine Classification. I have chosen 0.2% performance is greatly influenced by many
because this is normally the largest amount factors related to closed loop operation. This
of error we would leave when recalibrating a process is occurring continually during a
Class 1 type, testing machine. This value test. Therefore, the changing physical para-
may be as large as ±1.0% for a Class 1 meters being applied to the specimen as the
machine. specimen’s physical characteristics change
cause the system to react producing a
Repeatability is included in the estimate dynamic response.
for Uncorrected Error. Repeatability may be
treated separately depending on the calibra- Does this mean that in order to have a
tion or evaluation procedures used. defined level of confidence in the data pro-
duced for any test requires a dynamic cali-
So far we have examined uncertainty bration or verification? Possibly. But before
contributors related only to the static calibra- anyone should under take such an involved
tion and static performance of the testing task, there are a number of other things that
machine. This is an important point. Often should be investigated.
this is where the uncertainty analysis ends.
Currently accepted Material Testing Machine
calibration /verification procedures(11) allow Testing Machine Uncertainty
for the calibration of the system at low forces During Use
with certified deadweights. This essentially
means that the force applying apparatus of The measurement uncertainty contribu-
the testing system need not be turned on or tors in the section have the potential of mak-
running during the calibration. This will pro- ing all calibration sources of measurement
vide for very stable and repeatable calibra- uncertainty insignificant. Keep in mind that
tion data, but will not reflect any real world sources of uncertainty are combined in the
application. The problem is that very few real RSS method. This results in added weight
world tests are purely static in nature. It is my for the major contributors. A material testing
opinion that most tests performed with mod- machine ill suited for a particular test can
26 EUROLAB International Workshop: Investigation and Verification of Materials Testing Machines

contribute errors in force measurement and sources. I have designated those error
application well in excess of all other com- sources for which I have no objective evi-
bined uncertainties related to the testing dence as Not Available (N/A).
machine’s use.
Uncertainty Testing Machine During Use:
I do not have data to provide representa- Utmdu = v0.042 + 0.12 + 0.52 + 0.12 +
tive uncertainty values for all listed error 0.012 + 0.12 = 0.53%

Force Measuring and Application System Effects


Source Uncertainty Distribution
Drift .04% rectangular
Noise .1% rectangular
Resolution .5% rectangular
Stability (servo-hydraulic supply) N/A
Backlash (electro-mechanical) .1% rectangular

Environment
Source Uncertainty Distribution
Temperature .01% rectangular
Power Fluctuations N/A

Specimen Alignment
Source Uncertainty Distribution
Testing Machine and Grips N/A
Damage to the machine N/A

Application and Procedural Errors


Source Uncertainty Distribution
Errors due to system zeroing .1% rectangular
Specimen preparation N/A
Errors in reading displays .5% normal
Test Speed .2% to > 10% rectangular

I have not included the uncertainty contri- The combined Testing Machine
bution values for Errors in reading displays or uncertainty can be expressed:
for Test Speed because I have included a CUtm = v Usc2 + Ustm2 + Utmdu2 or
fairly large value for resolution. Depending on CUtm = v 0.262 + 0.282 + 0.532 = 0.65%
the type of evaluation and testing being per-
formed, one of the other sources of resolu- Expanded Testing Machine Uncertainty
(12)
NIS 3003 Edition 8 May tion error may become the dominate contrib- (k = 2) (12) can be expressed:
1995, ANSI/NCSL
utor resulting in a value less than or greater EUtm = 2 x CUtm or
Z540-2-1997
than 0.5%. 2 x 0.65 = 1.3% for a confidence level
of 95%
EUROLAB International Workshop: Investigation and Verification of Materials Testing Machines 27

Force Measuring and Application may not be economically feasible so it is an


System Effects area of potential error that should be investi-
gated.
Drift during use can be related to the sys-
tem’s inability to control well. The system Backlash in electro-mechanically driven
may need to wonder off the target force by a testing machines can influence testing
relatively large amount before the control results. The amount of backlash present can
signal makes an adjustment to correct the be assessed during calibration. Backlash
system. Drift can also occur due to system can also be minimized by causing the
devices changing relative to temperature. crosshead to advance past the point in
Creep in the system’s force transducer can which the test is to be started and then
add to this uncertainty value as well. adjusted to the start point of the test in the
intended direction of the test. This is a pro-
Noise can occur in the form of mechani- cedural issue and is only effective for unidi-
cal noise, electrical noise, or both. Mechan- rectional tests. If automated bi-directional or
ical noise to some degree is almost always cyclic testing is to be performed with these
present when the system is running. In mod- types of machines, the uncertainty contribu-
ern well designed and controlled systems tion should be determined and include in the
operating with in normal operating ranges of combined measurement uncertainty.
measurement and control, the total noise
due to electrical and mechanical influences
may be ±0.1% or less. Environment

Resolution error can be due to noise, data Temperature during a test may be signifi-
acquisition capabilities, and or test speed. If cantly different than the temperature during
resolution is evaluated as a factor of noise calibration. An evaluation of the effect of
during test, the uncertainty contributor for this difference should be performed. Where
resolution determined from the static calibra- applicable, test results may be corrected
tion analysis need not be summed in the due to temperature differences. Temperature
total combined uncertainty value. See Test changes during the test may also affect the
Speed below for additional explanation. test results. These gradients should be
known and included in the uncertainty analy-
Stability can be affected by the number sis.
of servo-hydraulic systems on a single
hydraulic supply. It is fairly easy to know A typical load cell temperature coeffi-
when the pump does not have enough flow cients specification.(13) (13) Interface, Inc. 1200

series load cell specificati-


or pressure to produce the end level forces
ons, 2000 product catalog
required. But it is not so easy to assess what Effect on Output - %/ºC Maximum: ±0.0015
happens to the specimen or system when Effect on Zero - %RO/ºC Maximum: ±0.0015
running long tests and numerous machines
on the same supply are starting and stopping We have found in our testing that as long
tests. The only way to know the effect is to as the load cell temperature remains at or
experimentally test this condition. I have not very near to the temperature at which the
included an uncertainty contribution value test is started, the error due to the tempera-
because it is highly subject to the type of ture being significantly different than the
testing and type of specimen /component temperature at the time of calibration is mini-
being tested. Ideally, there would be one mal. Because the system load cell is normal-
testing machine per supply or only one ly zeroed at the beginning of a test, if the
machine would be running at one time. This temperature then changes during the test,
28 EUROLAB International Workshop: Investigation and Verification of Materials Testing Machines

this is when the system force device will With all potential sources of measure-
contain errors when indicating force. I have ment uncertainty in this section thus far,
included an amount of uncertainty for this round robin testing between different testing
contributor that would reflect a 5 ºC change machines and if possible between different
in temperature from the start of a test. laboratories using adequate reference mate-
rials can provide evidence that the combined
Power fluctuations may affect testing measurement uncertainty due to these con-
machine operation and data integrity. If the tributors are minimized and with in adequate
power source meets recommended manu- control limits.
facturer’s specifications, no value for mea-
surement uncertainty need be included for
this condition. Application and Procedural Errors

These sources of measurement uncer-


Specimen Alignment tainty have the potential to be the most sig-
nificant sources of errors when performing
Good specimen alignment can be critical tests on materials and or components. These
to the life of the specimen being tested sources of measurement uncertainty are also
and thus important to the data characteriz- the most commonly overlooked sources
ing a particular material property. Testing when performing an uncertainty analysis.
Machines and Grips are manufactured to
apply and maintain good alignment. Again, it Common data acquisition zeroing errors
is difficult to put an uncertainty value to this can occur when the test operator arbitrarily
contributor because the value would be only zeros the data acquisition system at the start
significant for a specific type of specimen of a test. Careful attention and an under-
and test. Some manufacturers produce standing of the test conditions and testing
alignment devices that are easy to use and system configurations are important or sig-
can be adjusted with force applied. I have nificant offsets in the resultant data can be
found these to work well when good align- induced. This source of measurement uncer-
ment is critical. tainty is specific to the test and can have a
wide range of values. Fixturing, preloading,
Damage to the machine can occur at and backlash can have an effect on this
any time. Inspection of the system is critical component. I have included a value for mea-
to maintaining operational performance. A surement uncertainty equal to the amount
damaged machine can induce an out of commonly experienced when the system is
alignment condition. Poorly maintained sys- not zeroed after performing initial hysteresis
tems can lead to seal friction problems, cycling or if hysteresis cycling is not per-
which can result in loss of hydraulic pressure formed.
and oil leakage during cyclic testing. Servo-
Hydraulic systems require a clean oil supply Specimen preparation is critical to repeat-
in order to operate optimally. Most manufac- able and predictable results. It is my experi-
turers will offer oil sample testing. Poorly ence that most testing laboratories do a good
maintained or damaged electro-mechanical job of following established procedures when
testing machines may have excessive gear fabricating and preparing specimens. But,
play and or drive belts that are stretched. even relatively small dimensional machining
This can cause excessive backlash and start errors can cause large errors in specimen per-
up lag times. Assigning a value for measure- formance. Again this is a difficult measure-
ment uncertainty due to machine wear and ment uncertainty source to quantify. Lot test-
or damage is difficult. ing can often uncover problems of this type.
EUROLAB International Workshop: Investigation and Verification of Materials Testing Machines 29

Errors in reading displays are quickly pre-travel before force was applied. The brit-
becoming a small source of measurement tle nature of the polycarbonate specimen
uncertainty due to modern computer con- resulted in a test duration that lasted only 0.2
trolled testing machines. I still on occasion second. There were two basic problems with
have situations where the customers are try- this test. The test engineer assumed that the
ing to determine a force or data point on a crosshead would be moving at the expected
stress-strain curve below the first one inch speed when the failure occurred. Specimen
increment on a 10 inch chart recorder. Often failure was defined as the peak force at
times the potential error related to this prac- which the component physically came apart.
tice is neither assessed nor included in an
uncertainty analysis. Errors can easily equal
± 0.5%.
Crosshead Speed Ramp
20
Test Speed The largest errors I have
Speed (in./min)

experienced related to the testing of materi- 15


als and or components have been due to
10
inappropriate application of the testing
machine. Testing machines are expensive 5
and it is not unrealistic to expect users of
testing machines to use their systems for as 0
0 5 45 85 125 165 225
many different types of tests as possible.
Many times the system’s limitations are not milliseconds
well understood. Here is an example of a real
world situation I was involved in.
The graph shows that the crosshead was Graph 1 – Testing Machine
Crosshead Speed
A testing laboratory designed a test pro- moving at approximately 16 inches/minute
tocol to test a polycarbonate component just prior to specimen failure. The customer’s
used in the medical industry. The test engi- test protocol specified that the test speed
neers wanted to simulate a patient’s instan- have a tolerance of ±5% of desired speed.
taneous arm movement, such as a convul- The graph shows that the speed at specimen
sion or involuntary muscle reaction. This failure was 20% below the test criteria.
movement would result in applying force to
the component attached to the patient via The second mistake was to assume that
intravenous tubes. A Tensile Pull Test was the testing machine’s data acquisition sys-
designed to determine the force at which tem was adequate to capture the force indi-
the polycarbonate component would break cation at the time of specimen failure. The
under such conditions. A minimum force was customer expected that the error of peak
required to validate the component for con- force indication would be no greater than
tinued production. The test protocol was ±5%. This was also not the case. Table 3, Table 3 – Date Acquired
from Tensile Pull Test of
performed for validation of design as well shows the data acquired just prior to speci- Polycarbonate specimen at
as for continued production. Manufactured men failure. 20 inches/minute
samples were brought in periodically for lot
qualification testing.
Force (N) ∆ Force (N) %

The test was designed to run on an elec- 1919


tro-mechanical material testing machine. The 2106 187 8.9
system was selected to move the crosshead 2301 195 8.5
at 20 inches/minute. The specimen to be
2507 206 8.5
tested was fixtured so that there was zero
30 EUROLAB International Workshop: Investigation and Verification of Materials Testing Machines

By examining the data shown in Table 3, data fast enough. Few systems are capable
a couple of important things are apparent. of sample rates sufficient to meet the
The data in the ∆ Force (N) column shows requirements of this testing protocol. The
the difference in force between sequentially best way to run this test and acquire data
acquired data points. The difference between with in the accuracy desired is to simply slow
the first two data points is 187 N and the the test down. When the test was slowed to
force between the last two data points prior 2 inches/minute the data acquired averaged
it specimen failure is 206 N. Because the -0.43% with a maximum error of -1.22%
force is increasing between each acquired based on 12 tested samples. This of course
data point, we know that the crosshead is would not provide the instantaneous move-
still accelerating to the desired test speed. ment the design engineers had originally
The next data point if the specimen had not wanted to simulate.
failed would likely have been approximately
216 N or 8.6% from the previous data point. The equipment used to validate the test-
The system reported the specimen failure at ing protocol consisted of a certified traceable
2507 N. The specimen could have failed any- high-speed data acquisition system capable
where with in an 8.6% window greater than of 100,000 samples per second with 16 bits
2507 N. Data from the reference device used of resolution. The data acquisition system
to prove these findings recorded an average was connected to a certified force-proving
error in the reported peak force at failure of device fixtured in the load train during the
-7.22% with a maximum error recorded at test. I wrote custom software to use with the
-19.47%. This was based on a population of system in order to acquire and present the
60 tested samples. results.

It is important to note that this test result- An interesting side note to this scenario
ed in errors that where conservative in is that the customer would probably never
nature. The actual peak forces required to have become aware of this problem if they
cause the specimen to fail were much had not purchased a new machine for testing
greater than the test results would indicate. in the laboratory. They planned to run the
Therefore related to safety and liability this same tests on both machines to increase
was an overly safe test. testing efficiency and found that the new
machine produced data quite different than
This system was not capable of perform- their older machine. While I was there, I test-
ing the test with in the expected design ed their older machine as well and found that
specifications. There are a couple of things the results from that machine were consider-
that could have been done in order to run ably worse than the new machine. We deter-
this test closer to the design criteria. In order mined that the older machine was sampling
for the crosshead to be moving at the at 50 samples per second where as the new
desired speed when the event occurs, the machine was sampling at 200 samples per
system needs sufficient time to overcome second. Errors in peak force indication for
ramp up conditions. This could be accom- the older machine averaged -35% with a
plished by designing fixtures that let the maximum error of -45%. Neither machine
crosshead move a distance prior to the was sampling fast enough to produce data
application of force on the specimen. In with in the desire specification.
order to acquire test data with in the desired
±5% specification, the system’s data acqui-
sition system would need to sample much
faster. The system had sufficient static reso-
lution but dynamically it could not acquire
EUROLAB International Workshop: Investigation and Verification of Materials Testing Machines 31

Uncertainty of Test Results Recommendations

One method of determining the uncer- These recommendations are intended to


tainty of test results is to obtain the standard minimize measurement uncertainty and pro-
deviation of a series of tests performed with vide increased confidence in data produced
one particular set of control samples (Ucsr) by material testing machines.
on the same machine. The standard uncer-
tainty will include data scatter attributed to 1.) Once an uncertainty analysis is complete,
the samples. If reference material samples concentrate on reducing major sources
are available, tests can be run and the stan- of measurement uncertainty.
dard deviation derived (Urmsr). These values
are very dependent on the type of materials 2.) Know your testing machine’s capabilities.
and tests performed. The range of values Design your tests to operate with in the
can easily range from 0.1% to 1.0% and testing machine’s capabilities. Verify
greater. For the purpose of this exercise I will experimentally that all testing protocols
assign 1.0% to Ucsr and 0.5% to Urmsr. are operating with in expected specifica-
tions.
The combined measurement uncertainty
for the testing machine and testing results 3.) Examine the raw test data from your
can be derived. tests. Verify that the system is reporting
the correct results from the raw data.
Combined Uncertainty of Testing Results:
CUtr = v Usc2 + Ustm2 + Utmdu2 + Ucsr2 4.) Test reference materials in a number of
- Urmsr2 machines and compare results between
machines. Participate in round robin test-
CUtr = v 0.262 + 0.282 + 0.532 + 1.02 - 0.52 ing of reference materials between differ-
= 1.08% ent laboratories.

Expanded Uncertainty for the Test Results 5.) Keep the testing machine in optimal
(k = 2)(14): operating condition. Have your testing (14) NIS 3003, ANSI/NCSL

Z540-2-1997
EUtr = 2 x CUtr = 2.16% for a confidence machine on a scheduled maintenance
level of 95% program and have it serviced by trained
individuals.
Note: I have not included any value for
the uncertainty due to errors induced by 6.) Keep your testing machine in a calibrated
acceleration, system resonance, or other condition. Shorten calibration intervals
dynamically induced error sources. It is not based on historical calibration results
recommended that tests be performed where applicable.
when dynamically induce errors are present.
An evaluation of dynamically induced error
sources is beyond the scope of this paper.
32 EUROLAB International Workshop: Investigation and Verification of Materials Testing Machines

Bibliography

[1] ISO 10012-1: 1992 (E), Quality [5] ANSI/NCSL Z540-2-1997,


assurance requirements for mea- U.S. Guide to the Expression of
suring equipment – Uncertainty in Measurement.
Part 1: Metrological confirmation
system for measuring equipment. [6] ASTM E74-00a, Standard Practice
of Calibration of Force-Measuring
[2] ISO 376: 1999 (E), Metallic materi- Instruments for Verifying the Force
als – Calibration of force-proving Indication of Testing Machines.
instruments used for the verifica- ASTM Volume 03.01.
tion of uniaxial testing machines.
[7] ASTM E4-99, Standard Practices
[3] ISO 7500-1: 1999 (E), Metallic for Force Verification of Testing
materials – Verification of static Machines. ASTM Volume 03.01.
uniaxial testing machines –
Part 1: Tension/compression test- [8] EAL-G22, Uncertainty of
ing machines – Verification and Calibration Results in Force
calibration of the force-measuring Measurements
system.
[9] Interface, Inc. Product Catalog
[4] NIS 3003 Edition 8: May 1995, 2000.
The Expression of Uncertainty and
Confidence in Measurement for
Calibrations.

You might also like