w \.
ASME MEASUREMENT UNCERTAINTY
AB. 8. Abernethy >
Pratt w Wittney Atrera?t Company
Henber ASHE
RP, Benedict
Westinghoure Clectr ie, Carporatiun
Feltow ASKE
RB. Dowde}
University of Mhnde land
None ASK
awsTaacT
The purpose of this paper is to introduce the new ASME
neasurenent uncertainty mothnialagy whtrn fc the mack
Tor" two new RIME/ANSE standards ond the RSME short
course of the sate mane. Some’ background and history
that Ted to the selection of this methodology are.
discussed as well as {ts application tn current SAE,
TSA, JAWIAF, RC, USAF, NATO ano 150 standaras,
Aociments ahd short courses. “This ASME sethodology 1s
aptly Becoming the national and international
standard.
ancxeRouNo
The accuracy of test results has always concerned
engineers and scientists, but for decades this subject
nis been plagues oy conttoversy,. argument, ‘contuston
tng even Snotion.. The absence of an uncertainty
Calculation standard nade stanificant conparison of
test resuTts between facilities, companies. and
attempts, #. HI'tu oF NBS relates the following Ci}és
“Dan Jonason, an old timer at the Bureau, told me.
this story. "tn the 1930's, P. He Myers at AUS and
his col sagues were Studying the specific heat oF
Simonta. After several years of hard work, they
Finally’ arrived ata value and reported the result
Ing paper. “Towars the end of the paper, Wyere
“we think our reported value ts good to one part in
10,000; we are willing £9 Bet our own Money at even
dic that it te"earract ta ten naete dn fi
Fartnermore, if Oy" any ‘chance our value 1 shown to
bein error’sy more than one part in 1,000, we are
Drepared to eat our apparatus and drink the
Sowonta”
* dunbers In brackets designate references at the end
wistoay
In the research that, led to the JAMAF (formerly
renee [iva ete usar [5] membwehy $ poner fat
Statistical tool, Monte Carlo simulation, was used to
Select the best ietnods fron the many avatTabie.. J.
Rosenblatt, h. i. Ky and d. M. Caneron of MBS provided
‘excel Lent constructive criticisn of these docunents
Snd have continued to support industry in Enis,
erfort.. The references to the NUS publications are
particularly recommended to the reader. [ly 4y 5, 6).
Oy we vave seventies, che unly mayor argument ena
FPoadined was over how'to combine’ the bias error Tint t
itn the precision error. Addition of the. two
Components is recoomended in [2, 3, & T, 12, 13 and
Ist. "Combination by the root-sin-iquare’sethad ¥<
recomended in (8,-9, 10, 15, and 1a). This ergunent
Could not be solved Conpietely by Nonte Carlo
Simulation a5. 1t 15 Vargely a aatter of opinions
However, these Simulations aided significantly in
eratuacing ete Seatistical craractertsties OF tne ‘40
Uncertainty intervals. The argument. as to how tO
Conbine bias and precision errors raged over many
Comittees. in several societies, and most. participants
Nelfeved it mould never bo cettied.. & compronise was
suggested by the Wis group (6) in late 1980. Tt was
Suggested that (1) if the bias and precision
Components are propagated separately fran the
measurements to the Final test result and (2) the
fethod of combination 1s clearly stateds, then el ther
the addition or root-sun-square cethod should be
accopted ag it 1s the Tast step in the caTeulation and
Ean easily de undone. Shortly thereafter, the ASME,
AMowing the analyst to decide and state which
lncertataty model (00 or ASS) was to De usedCURRENT ACTIVITIES
ASME. The two ASME comitters are:
[AUSI/ASHE PTC 19.1 - 1983 - Heasurenent
Uncertainty =
‘MIST /ASME.WFC-2-1983 ~ Uncertainties in
Flow Measurement —
In addition, the ASHE Short Course on Measuring
Uncertainty is scheduted for the 1983 ASKE-HAN,
‘The status of these ANSI/ASME documents 4s
described in Tater sections of tris paper.
SAL, Comittee £33 on “Aircraft. In-Flight
pulsion Measurenent and Uncertainty," is
drafting an SAE Aerospace Infornation Report (AIR
Be) tieleg, acFtahe Trust Measure
Uncertainty, which they hope to distribute for
‘naustry review in 1984, This docunent uses the
sane uncertainty methodology as that of tis
acer. The activities of this connittee are
Sescribed in {IL}.
SA, The Instrument Society of Anerica provides
2 slort course titled, Test Measuresent Accuracy,
{it the International instrumentation Symposium
fad other locations several tines a year. This
course 15 identical to the ASME Short Course.
TSH alzo has formed a Measurement Uncertainty
Comittee to encourage and pronote the use of
Imeasurenent uncertainty analysis. The United
States Air Force Handbook (3] as been reprinted
150, 150 TC30 5¢9 approved the method described
herein at theiy meeting in Leningrad in May 1982
and roquczted a revision of the ext sting world
Standard 190 5168 [11]. The second draft was
reviewed at their recent meeting in kashington,
D.C. at the National Bureau of Standards in
November 1902,
MIOAP. The British Minfstry Industry Orag
‘Analysis Panel published their report,
‘Agaroyraph 2375-10 1979 Co]. Una Joint
eeting held with SAE Committee £33 in England in
Yay 1982, the uncertainty methodology was
coordinated between these to groups.
ASG. The AUSI Committee Z11 has commissioned an
{SOC Kriting Group on Calibration Assurance.
This Writing Group 1s drafting a national
Standard on assuring the quality of calibration
C15]. Although this standard treats only
altbration error, it is consistent with the
rethodology recomended herein,
7. MIC. At the Idaho National Engineering
Laboratory, a report titled, Sentscale
Uncertainty Report: Methodology [14T, has deen
written Tor te United States Huclear Regulatory
Comission. This document uses the uncertainty.
‘sethodology descrited herein. Nuclear Material
Control, Nase Calibration Techniques, AIST
‘MI5.18-1975, also is cons{stent with’ the
recomended nethodology.
8. NATO. NATO AGARD PEP 15 conmittee on Uniform
Engine Testing 15 conducting an interfactTity
test of two Jet engines at NASA-LEVIS, USAF KEDC,
USN MAPC, Britain's NGTE, France's. SACLAY
FaciTity and a Turkish facility. This comittee,
selected the recomended uncertainty methodology
as their standard for this progran at their
reeting in Toulouse, France in May 1981.
9. GRC. The Coordinating Research Council has
“ecided to re-evaluate the test data fron thelr
Atlantic City test progran on engine exhaust
fenissions usina the recomended uncertainty
‘nethosotoay.
BRIEF DESCRIPTION OF METHODOLOGY
Measurenent Error
It 15 4 well-accepted principle in engineering that
aNl seasurenants have errors (é]~ These errors. are
the differences between the meaSurenents and the true
value (see Figure 1). Furthermore, the total error ts
usually expressed in terms of two Components: a Fixed
(eas) corer (p Vy onda randaw (prectaten) error
(og) such that
grate w
ave VALUE a
roratation
i
vate
Fue 1
MEASUREMENT ea908Precision Index
The precision error is determined by taking N repeated
‘neasurements from the paraneter population, the
Characteristics of which can be approxinated by the
precision index (5)'defined by the familiar
we
o, - 04
vere T is the average value of X.
‘The precisfon index of the average of a set of
rmeasurenents {5 always Tess than that of an individual
rmeasurenent according to
Se 7 o
Bias Error
‘The bias error 1s the systematic error which 1s,
considered to romain constant during a given test.
Thus, in repeated measurenents of a given set, each
seasuranent har the sana Bias. Thora He 0
Statistical equation, a5 (2) or (3), to define the
bias Tinit, 8. Instead, 1t must be estimated, and
this is not an easy matter since the true value 1s not
known. Cal Brations helpy 22 docs a conparizon of
Imeasurenents by independent methods, but in general
the estimate of bias must be based on Judgnent.
Coubining Errors
Errors arise from many sources. These are divided
arbitrarily into three categories: calibration
errors, dale acquisition errors, and date reWuctton
terrors: For each of these sources of error there wil)
be Bias and precision components.
To obtain te precision of a given paranever (11ke
tenperature, pressure, or flow rate), the root sua
‘square (RSS) nethod 4% used to conbine the precision
‘indices fron the K sources of error. Thus
saCsteste este . a
Similarly, the bias of a given paraneter 1s given by
se Catt at ee geT 2 st
Uncertainty of a Parameter
If a single nunber (U) is needed to express a
reasonable limit of error for a given paraneter, then
same model for combining the bias and precision errors
‘must be adopted, where’ the interval
weu GC)
represents a band within which the true value of the
paraneter is expected to Tie, fora specified
overage.
Unite no rigorous confidence evel can be accnctatad
with the uncertainty (U}, coverages analogous to the
85% and 99% confidence levels can be given for the two
reconnended uncertainty models. Thus
9% coverage, (7)
we
and Uggs = (8% + (155141 provides 3)
Uy = 8+ ESE provides
958 coverage.
The Student t value 1s a function of the degrees of
Freedom (y ) used in calculating S-. For large
samples, (i.e, 1P30), t is set eqlal to 2, otherwise
according to .
a aeae
cd Sut a
a a
AM, oO
Bo Sy
sho gO
M5
sere Si represents the precision tnsies of he
‘degrees of Freedom oF these sane error’ sources.
Uncertainty of a Result
Errors in measurements of various paraneters (P) are
propagated into a derived result (r) through the
functional relattonship between the result and fts
‘independent’ paraneters. The relationship provides the
sensitivity factors (0,), wich indicate the error
propagated to the result because of unit error in the
parsneter. Thus if
= FPL, Py eee ND 10)
where J 1s the nunber of paraneters involved,
then
-z
4- Ro. ayne tas ana precision errors of the paraneters are
kept separate until the Tast step of computing the
uncertainty of a resuTt. Thus, the precision index of
a result 15 given by
tb ts 83 1? ne
en
and the bias Vintt of a resute 4s given by
1 12
CE ty wea, 03)
4
The uncertainty of a result (5 agafn afven by the two
models according’ to
Y= + tS, 8 om an
300
and
12
uC a2 + tts)2 2 @ on, as)
aS
‘The Student t value 1s a function of the degrees of
freedon used in calculating S;. For large samples of
all paraneters, (i.e., 130),"E 1s set equal to 2,
Gthersise the lelch-Satterthwat te formula 1s used’ to
provige v, according to
ro)
ASME PTC 19.1, MEASUREMENT UNCERTAINTY
This committee was formed in 1979 to provide the
Supplement’ on which to base Measurenent Uncertainty
Analyses. This was for the use of the various Code
and Suppienent writing committees.
‘This coomtttee has endorsed and contributed to the
‘ethodology of this paper, and has Just completed a
draft for the PIC Board and Industry approval [16].
‘The docunent includes a nomenclature and a glossary of
terms that are in agreement with the various
International Standards. A detatled review of the
Imethods of this paper 5 Included, ay fs @ strony
Section on Applied Considerations, This latter
includes: miltiple test uncertainty, long versus
short term tests, comparative versus absolute tests,
Spacial variations, ovtl er treatment, regression
uncertainty, weighting method, pre- and post-test
analyses, and number of measurenents required. A
step-by-step calculation procedure is given, as well
45 workea-our examples applying the method.
AIL in all, we expect to satisfy the PTC requirements
for an authoritative docunent on measurenent
uncertainty that is casily understood and applied.
ASME EC, FLUIO FLOW MEASUREMENT UNCERTAINTY
‘he ASHE Standards Consittee on the Heasurenent of
Flutd Plow tn Cloced Conduite (HEC) war formed tn 1972
as a result of the recognition by those actively
working in the field that there was a need for a
Single national standard on this subject. In most
Vestorn European countries, ational standards. on flow
measurement have been in use for mary years. These
fare usually promulgated by goverment supported
agencies such as the British Standards Institute in
Germany etc. "The First International Standard on Flow
Measurement 1S0/RS41 was publishes in January 1967.
This country has had no national standerd, but many
authoritative documents on flow neasurenent existed
such as the ASHE Report "Fluid Meters, Their Theory
fand Application’ (17), PTC 19.5 0n Flow Measurement,
‘cA Report #3 on Gas Flow Measurement, ele. For Une
Imost part, these documents were in agreesent on their
methodology, coefficient values, required upstreae
‘engths, and calculation procedures.
This was not true on the international scene, and the
‘initial 190 document contained many compromises
between USA procedures and those in use throughout
Western Curope- Differences Im required upstream
lengths and coefficient values have not yet been
resolved.
‘AN of uese publ teattons, those within the WA, tne
European national standards, and the international
Standards, address the question of the accuracy of a
low measurement. However, each docunent created sts
‘om procedures for estimating the uncertainty ang
values given were based on hunan judgeent usually
biased by the individuals involved.
‘we Firse puov ication or a stanaara aevotea entirely
to the estfmatfon of uncertainty of a flow rate
‘eeasurenent mas. 150/015 5168 published in 1976,
Recognizing the inportance of this subject matter, the
ASHE. Standards Committee MFC set up its first,
Subcommittee, {.e., SCI with the charge to prepare a
USA standara’‘on ‘Uncertainties in Flow Measurenent, It
has taken many years, 10 to be exact, and much effort
by the people involved, to produce the first ANSI/ASHE
EC ~ 2H Standard on Uncertainties in Flow Neasurenent
which was published in 1983 C18]. The methodology
follows that of the preceding section of this paper
‘and should form the basis of some further, mare
applied, or working documents on flow measurenent.,Soma
Engineering judgnent and experience 1s stfl1 required
When estimating bias or systematic errors but this 15
Clearly statea tn [1s] and all parties t0 a contract
an agree beforehand to the values that should be
used.
Unfortunately we are stil] left with many unanswered
(questions that ust be resolved fn the not-too-di stant.
future.” For instance: How do ve interpret. statenents
by the manufacturers of industrial Instrunentation
that claim a device to be “accurate to within #0.5% of
full scale"? That 1s: How such of this is blas error
and how such should be attributed to random or
precision error? Sintlarly, whan dealing with mater
Coefficients: How do we interpret values given in
LIM] for the 20 tolerance on the discharge
coefficient? It is necessary that the concepts of
thie paper be adnnted thrnaghout ATT eagmante af
industry and one day we will have a uniform,
unanbiguous method of estimating the uncertainty of
not only a flow measurenent, but measuresents of all
REFERENCES
(17 HM ku, umceRTAINTY 18 ns's aUSiMEss,
Presented to the NBS Advisory Panel, October 9,
1975.
v2
Abernethy, et a ToRPG HANDBOOK FOR
ESTIMATING THE UNCERTAINTY In| MEASUREMENTS. MADE
WITH LIQUID PROPELLANT ROCKET ENGINE SYSTEMS,
SNA (formerly ICRPG) Perforsance
Standardization orking Group Report CPIA Wo, 100
(40 851 127), April 1969,
[3]. R. Bs Abernethy, et al, HANDBOOK UNCERTAINTY IN
hs TURBINE MEASUREMENTS, USAF Arnold Engineering
Development. Center, Report AEDC-TR-73-5, February
1973.
TOD te we May Cutan PRECESION REASUKEREN AND
CALIBRATION, NBS'Special PusTication 300, Volune
1, 1969.
15)
Rosenblatt and C. Ha Spiegelman,
Novenber 1981. a
6] was POSTSCRIPT TO SPECIAL PUBLICATION 300,
[71 MW. 0. Hersey, A DEVELOPMENT OF THE THEORY OF
ERRORS WITH REFERENCE TD ECONOMY OF TIME, British
Association for the Advancenent of Service, 1913,
Reprinted by ChurchiT] Eisenhart, M85, Journal of
Research, 1965.
[8] 5. 0. Kline, Fe As MeClintock, DESCRIBING
UICERTAINTIES iN SINGLE-SANPLE EXPERIMENTS,
MECHANICAL Engineering, Volume 75, 1953.
[9] MIOKP Study Group, GUIDE TO IN-FLIGHT TaRUST
YEASUREHENT OF TURDOJETS AND. TURBOFANS,
AGARDOGRAPH AG-237, January 1979.
[10] 150 5168, MEASUREMENT OF FLUID FLOM -- ESTIMATION
(OF UNCERTAINTY OF A FLOW-RATE MEASUREMENT, 1976,
[11D &. @. Aherneehy, SAE IN-FLIGHT panpuLsioN
MEASUREMENT COMMITTEE E93: ITS LIFE AND MORK,
SAE in Aerospace Engineering Volume 1, No. sy
uly 1981.
[12] R. B. Abernethy, et al, Instrument Society of
‘America MEASUREMENT UNCERTAINTY HANDBOOK,” ISBN:
87664-483-3 Revised 1980,
(13) RW. Golden, SEMISCALE UNCERTAINTY REPORT:
NETHOOOLOGY, "prepared for U.S. Nuclear Regulatory
Commission, MUREG/CR-2459, Sept. 1982.
[14] R. 8. Fe Schumacher, SYSTEMATIC. MEASUREMENT
ERRORS, 'ASQE Journal of Quality Technology,
January 1981.
£15] ASQC writing Group Oraft, AMERICAN NATIONAL
‘STANDARD FOR CALTERATION SYSTEMS, 1963.
(16) ASL/ASIE PTC 19.1, 1903, NEASURENENT
UNCERTAINTY,
(17] Fluid Neters, Tetr Theory and Application, 6th
Eattion, 1971, puolisned by ASME.
[18] ANSI/ASHE NFC-2, 1983, MEASUREMENT UNCERTAINTY
FOR FLUID FLOM Ii CLOSED CONDUITS.