Anda di halaman 1dari 10

KOCE S S M E A S U R E M E N T

H("%ccurate
isAccurate?
P a r t i
WH A T'S A CCURA TEFOR ONE A PPLICA TION
MA Y BE A PPROXIMA TE FOR A NOTH ER.
WH EN A SSESSING A CCURA CY, TH E FIRST
STEP IS TOUNDERSTA ND TH E TERMS
B y Wi lli a m Most i a
William LMostia, Jr., P.E., is a
senior process control and
c ompu t i ng e ngi ne e r wi t h
A moc o Cor p.'s Wor ldwi de
Engineering &Construction
divisionin Houston.
Thi s a r t i cle is the f i r s t i n a t hr ee-pa r t ser i es. Her e, P a r t I exa mi nes t he concept s
a nd t er mi no lo gy used t o defi ne a ccur a cy of pr ocess mea sur ement s. In July , P o r t II
wi ll t ell ho w t o co mbi ne er r o r s wi t hi n a n i ns t r ument o r sy st em t o pr o vi de a n est i -
ma t e o f t he t o t a l er r or . In Augus t , P a r t III wi ll pr o vi de a met ho d t o det er mi ne ho w
a ccur a t e a n i ns t r ument o r sy st em mus t be t o per fo r m i t s speci fi ed f unct i o n.
T
he i mpor t a nce of a ccur a t e mea sur ement s i s obvi ous t o t hose
who wor k i n a pla nt wher e compensa t i onpa y a nd bonus-
esi s t i ed t o pla nt per fo r ma nce. The t hi ckness o f t hei r pa y
envelopes i s det er mi ned by how much pr o duct i s pr oduced,
a nd wha t i t t o o k t o pr o duce i t . Oper a t o r s , engi neer s , a nd t echni -
ci a ns i n t hose pla nt s ha ve di r ect i ncent i ve t o ma ke sur e di ei r i nst r u-
ment s a r e gi vi ng t he mo s t a ccur a t e r ea di ngs po s s i ble.
In o t her pla nt s , a ccur a cy ma y seem less i mpo r t a nt . Who ca r es
i f t he r ea di ngs a r en't exa ct , a s lo ng a s t her e's eno ugh us a ble pr o d-
uct t o keep t he bo s s o f f y o ur ba ck? If y o u (a nd y o ur bo s s ) a r e
lucky , y ou do.
Thes e da y s , i ncr ea s ed co mpet i t i o n a nd go v er nment r egula -
t i o ns ha ve bo o s t ed t he dema nds f o r i mpr o ved o per a t i ng ef f i -
ci ency , bus i nes s uni t a cco unt a bi li t y , co s t lea der s hi p, a nd qua li -
t y cer t i f i ca t i o ns . The a ccur a cy o f mea s ur ement a nd co nt r o l s y s -
t ems i s o f gr ea t er co ncer n. Ext ens i ve a ppli ca t i o ns o f co mput er s ,
da t a co llect i o n f a ci li t i es , a nd da t a ba s es a r e r ely i ng o n a ccur a t e
mea s ur ement s .
Accur a t ely mea s ur i ng a ll t he pr ocess ma t er i a l a nd ener gy flo ws
helps eli mi na t e wa s t e, i mpr ove o per a t i ng ef f i ci enci es , a nd r educe
cost s. B usi nesses a r e put t i ng s t r i ct er a ccount a bi li t i es a t lower lev-
els i n bus i nes s uni t s t ha t r equi r e a ccur a t e i nt er na l a cco unt a bi li -
t y f o r t he uni t a s well a s f o r i nt er co mpa ny a nd i nt r a co mpa ny
cus t o dy t r a ns f er .
ISO 9000 cer t i f i ca t i o n i s o f i ncr ea s i ng i mpo r t a nce i n t he pr o -
ANACCURATEGLOSSARY
Accuracy: The degree of conformity of anIndi-
cated value to a recognized accepted stan-
dardvalue'.Accuracy for Instruments Is nor-
mally statedInterms of error ( .05% of upper
range value [URV], 1% of span,0.5% of read-
Ing,3/4 degree, etc.).
Accuracy canalso be stated Interms of bias
andprecisionerrors.34 Looking at Figure 1,the
shift of the bullet holes from the bullseye Is
the bias andthe tightness of the bullet pat-
ternIs the precision error.
Bias may be knownor unknown.Anexample
of a known bias Is the deviation of a calibra-
tionstandardfrom a National Institute of Stan-
dards andTechnology (NISD/Natlonal Bureau
of Standards (NBS) reference. Large known
bias es are normally calibrated out. Sma ll
knownbiases are normally compensatedout.
Examples of unknownbiases Include human
error, installationeffects , environmental dis-
turbances,etc.2
Precisionerrors are consideredstatistical-
ly random. They canbe statedas the product
of the measurement's standarddeviationand
Student Tdistribution, whichwill provide an
error specificationto the 95% confidence level.
Also random are errors specified for trans-
mitters, calculationdevices,constant uncer-
tainty, recorders , input/output devices , etc.
Bias andprecis ionerrors canbe individually
c ombi ne d u s i ng t he r oot s u m s qu a r e ( RSS)
method, then the bias andprecision errors
canbe combined:
$ + ; ..
E= ( B+P)
where:
E tota l probable error,
B total probable bias e r r or s ,
b - bias errors,
P= total probable precis ionerror, and
p = precis ion errors.
Absolute accuracy: Howclose a measurement is
to the NIST/NBSstandard(the "golden ruler").
The accuracy traceabllity pyramid is s hownin
Figured
Conformity: The maximum deviationof a callbra-
j UME/ 1 996 CONTROL
tlon curve (average of upscale and downscale
readings) from a specified characteristic curve.'
Conformity can be independent (best fit), zero,
or terminal based. This spec is commonly used
as a measure of how close an instrument con-
verts a non-linear input signal to a linear output
signal. This error specification is typically seen
in temperature instruments.
Deadband: The range through which an input
signal may be varied upon reversal of direc-
tion without initiating an observable change
in the output signal.1
Drift or stability error: The undesired change In
output over a specified period of time for a con-
stant input under specified reference operating
conditions.1 Drift error is influenced by envi-
ronmental exposure and can be significant (or
greater than the reference error), but It can be
controlled by periodic calibration.
Dynamic error The error resulting from the dif-
ference between the reading of an instrument
and the actual value during a change in the actu-
al value. Instrument damping contributes to this
error as does measurement and transport
deadtime, and the significance depends on the
process time constant Dynamic error must be
considered when designing safety systemsif
your instrumentation system cannot measure
a developing hazardous condition In time for the
safety system to react, the-process may not get
to a safe state in a timely manner.
EMLUFI errors: Errors due to electromagnetic or
radio frequency interference.
Rlter error Error caused by the Improper appli-
cation of a filter on the signal. This error can
also be caused by improper settings In exception
reporting and compression algorithms.
Hysteresis: The dependence of the output, for a
given excursion of the input, upon the history
of prior excursions and the direction of the cur-
rent traverse.1
Influence errors: Errors due to operating con-
ditions deviating from base or reference con-
ditions. Typically specified as effects on the
zero and span, these errors reflect the instru-
BIAS VS. PRECISION
Pmaiw
THE SHIFT OF THE BULLET HOLES FROM THE BULLSEYE IS THE
BIAS ERROR AND THE TIGHTNESS OF THE BULLET PATTERN E THE
cess industries, and instrument accuracy
is an important part of that. The govern-
ment is putting more and more regula-
tions on the process industries, many requir-
ing more accurate measurements and data
collection.
So accuracy is important. But what is accu-
racy? The language of accuracy is not universal,
and any discussion depends on a common under-
standing of terminology. (For definitions of
commonly used terms, see sidebar.)
A b s o l u t e A c c u r ac y o r ' R e p e at ab il it y ?
By definition, all accuracy is relative: how accu-
rate a measurement is compared to a standard.
When discussing the error of an instrument
or system, we need to determine what
form of accuracy we need for a particular
function. Absolute accuracy refers to how
close a measurement is in relation to a PREC ISION.
traceable standard (see the traceability pyra-
mid in Figure 2).
"Repeatability," on the other hand, refers to how accurately a measurement can be
duplicated or repeated. (The term "repeatability" in this context is the common field
usage and not the ISA definition of the term. The common field usage is essentially
the same as the ISA term "reproducibility;" that is, the combination of linearity, repeata-
bility, hysteresis, and drift.)
If it is important that you make a measurement in reference to an absolute value,
we are talking about absolute accuracy. When most people talk about accuracy, they
are talking about absolute accuracy. In order to have absolute accuracy for your
measurement, you must have traceability from your measuring device to the Nation-
al Institute of Standards and Technology (NIST)/ National Bureau of Standards (NBS)
reference standards (the "golden rul ers"). The accuracy of your measurements is
directly dependent on the accuracy of your calibrators, which is directly related to
the care and feeding of your calibrators, the calibrator' s calibration cycle, and the
traceability of your calibrators.
In a process plant where ambient and process conditions can vary substantially from
the reference conditions, maintaining accuracy can be a daunting tasks. C alibration
cycle and methods, instrument location, instrument selection, maintenance, record-
keeping, and training all become important issues in maintaining your instrument's accu-
racy. A formal calibration program is the only way to ensure the accuracy of your
instruments. This is essential for achieving and maintaining an ISO 9000 certification.
In days past, and probably today in some plants, it was not uncommon for an operator
to control a flow to so many "roots" or some other variable to so many divisions. Here die
concern is how repeatable the measurement isif we are controlling to seven roots today,
we want seven roots tomorrow to be die same thing. We want the instrument to provide
the same value each time for the same process and operating conditions.
Many controllers diat have relatively crude setpoints, such as field pneumatic controllers
and HVAC thermostats, specify "repeatability." The object is to maintain an acceptable
setpoint, with little concern about the absolute value.
More critical applications such as laboratories and research facilities often use cali-
(continued onpS2)
C ONTROL J UNE/ 1 9 9 6
( c o n t i n u e d f r o m p 5 i )
br at i o n c u r ve s. The ac c u r ac y o f m e asu r e m e n t o r c o n t r o l i s r e lat e d t o a p ar t i c u lar r e ad -
i n g , whi c h i s t he n t r an slat e d t o an abso lu t e ac c u r ac y valu e u si n g a c ali br at i o n c u r ve . Fo r
t hi s t yp e o f m e asu r e m e n t , we ar e ag ai n t alki n g abo u t r e p e at abi li t y. Fo r d e vic e s whe r e
i n f lu e n c e e r r o r s ar e m i n i m i ze d an d t he d r i f t e r r o r i s c o n t r o lle d by t he c ali br at i o n c yc le ,
t hi s m e t ho d c an r e ac h a hi g he r le ve l o f ac c u r ac y.
TRAGEABILITY PYRAMID
A B S O L U TE A C C U R A C Y R E FE R S TO H O W C L O S E A M E A S U R E M E N T is
TO A TR A C E A B L E S TA N DA R D. S U C H .A S N IS T/N B S S TA N DA R DS .
Error Specifications
M an u f ac t u r e r s sp e c i f y e r r o r li m i t s f o r an i n st r u m e n t .
The se ar e n o t t he ac t u al e r r o r s t hat a p ar t i c u lar i n st r u -
m e n t wi ll have , bu t r at he r , t he li m i t s o f t he e r r o r t hat
t he i n st r u m e n t c o u ld have . O n an i n d i vi d u al basi s,
a g i ve n i n st r u m e n t m ay be able t o be c ali br at e d t o a
hi g he r ac c u r ac y t han i t s sp e c i f i c at i o n , an d a g r o u p
o f t he sam e i n st r u m e n t s wi ll f all wi t hi n t he e r r o r
sp e c i f i c at i o n .
If a m an u f ac t u r e r st at e s an e r r o r sp e c i f i c at i o n , i t i s
g e n e r ally t r u e t o wi t hi n t he ve n d o r 's t e st i n g m e t ho d -
o lo g y. The u se r sho u ld qu e st i o n an y m an u f ac t u r e r who
d o e s n o t g i ve an e r r o r sp e c i f i c at i o n . Y o u m ay f i n d t hat
t he r e i s a g o o d r e aso n t he sp e c i f i c at i o n was le f t o f f .
B u t m an u f ac t u r e r s d o n o t t yp i c ally g i ve ju st o n e
e r r o r sp e c i f i c at i o n . In st e ad , t he y g i ve m u lt i p le
sp e c if ic atio n s an d so m e t i m e s i n d i f f e r e n t ways.
Thi s i s be c au se t he e r r o r wi ll t yp i c ally var y
whe n am bi e n t an d p r o c e ss c o n d i t i o n s
var y f r o m t he r e f e r e n c e c o n d i t i o n s
whe r e t he i n st r u m e n t i s c ali br at e d , an d
t he d i f f e r e n t e r r o r sp e c i f i c at i o n s allo w t he
u se r t o d e t e r m i n e t he p r o bable e r r o r at
o t he r c o n d i t i o n s. S o i n r e ali t y, t he se e r r o r
sp e c i f i c at i o n s d e f i n e an e r r o r e n ve lo p e . A n e xam p le o f t he e r r o r vs. am bi e n t t e m p e r a-
t u r e e n ve lo p e f o r a g e n e r i c t r an sm i t t e r i s g i ve n i n Fi g u r e 3.
Not All Error Specifications Are Created Equal
E r r o r s ar e sp e c i f i e d i n a n u m be r o f d i f f e r e n t ways. In o r d e r t o c o m p ar e an d c o m bi n e
e r r o r sp e c i f i c at i o n s, t he y m u st all be o f t he sam e t yp e . S o m e o f t he t yp i c al e r r o r sp e c i f i -
c at i o n s ar e :
0.2% o f calibrated span inclu ding the co mbined effect s o f linearity, hyst eresis, and repeatability;
10.2% o f u pper range limit ( U R L ) six mo nths fro m calibratio n;
0.1% o f calibrat ed span o r u pper range valu e ( U R V) , whichever is greater;
10.5% o f span per 100'F change;
0.75% o f reading;
1*F;
1/2 co u nt; and
1 least significant digit ( L SD) .
Tr an sm i t t e r r e f e r e n c e ac c u r ac y i s t yp i c ally r at e d i n p e r c e n t o f sp an o r U R V whi le
p r i m ar y m e asu r i n g e le m e n t s, su c h as o r i f i c e p lat e s, t u r bi n e m e t e r s, an d t he r m o c o u -
p le s, ar e r at e d i n p e r c e n t o f r e ad i n g o r ac t u al m e asu r e m e n t e r r o r . Fo r a t r an sd u c e r t hat
i s c o n n e c t e d t o a t he r m o c o u p le , t he e r r o r sp e c i f i c at i o n s ar e n o t t he sam e t he f o r m e r
i s t yp i c ally i n p e r c e n t o f sp an o r U R V whi le t he lat t e r i s i n p e r c e n t o f r e ad i n g . In o r d e r
t o c o m bi n e t he se e r r o r s, t he y m u st be t he sam e t yp e .
( c o n t i n u e d o n p 5 4 )
merit's capacity to compensate for variations
in operating conditions. Influence errors are
often significant contributors to the overall
error of an Instrument
Inherent errors: Errors inherent in an instru-
ment at referen ce conditions. These are due
t o t he i n he re n t me c ha n ic a l a n d e le c t ric a l
design and manufacturing of the instrument
linearity. The deviation of the calibration curve
from a straight line. Linearity Is normally spec-
ified in relation to the location of the straight
line in relation to the calibration points: Inde-
p e n de n t (be s t straight line f it ), ze ro-ba s e d
(straight line between the zero calibration point
and the 10 0 % p oin t), or te rmin a l-ba s e d
'* (straight line between the zero and the
.," 10 0 % calibration points). Devices are
;.; normally calibrated at zero and at their
upper range value (URV)the terminal
points. It is, however, possible to calibrate
a device at other points to get a better fit
ISA Standard S5 1.1 provides a good descrip-
tion of linearity.
Moun tin g p os i t i on e f f e c t : The e f f e c t on t he
instrument's calibration due to mounting posi-
tion. This Is t yp ic a lly a ze ro s hift e rror that
can be calibrated out a fte r Installation.
Overrange influence error Error resulting fro m
the overranging of the instrument after instal-
lation. This is normallya zero shift error.
Power supply effect The effect on accuracydue
to a shift in po wer supplyvoltage. This erro r
cou ld also applyto the air supplypressure for
a pneumatic Instrument
R eference accuracy: Accuracy, typicallyin percent
of calibrated span, specified bythe vendor at a
reference temperature, barometric pressure,
static pressu re, etc. This accu racyspecifica-
tion mayinclu de the co mbined effect s of lin-
earity, hysteresis, and repeatability. R eference
accuracyalso maybe stated in terms of mo de
of operation: analog, digital, or hybrid.
Re f e re n c e jun c t ion c omp e n s a tion a c c ura c y: The
a c c ura c y of t he c old jun c tion c omp e n s a t i on
for the rmoc oup le te mp e ra ture transmitters.
J U N E /I 996 C O N TR O L
( c o n t i n u e d f r o m p S 2 )
Di gi t al d e vi c e e r r o r s ar e u su ally r e so lu t i o n e r r o r s whi c h ar e r e lat e d t o c o n ve r si o n
e r r o r s, r o u n d o f f e r r o r s, an d n u m e r i c al pr e c i si o n . Fo r e xam ple , a 1 2 -bi t r e so lu t i o n
i n pu t d e vi c e r e so lve s t he si gn al i n t o 0-4,095 c o u n t s. It c an n o t r e so lve t o le ss t han 1 /2
c o u n t o u t o f 4,095 c o u n t s.
Ro u n d o f f e r r o r o c c u r s whe n a d i gi t al d e vi c e r o u n d s o f f a par t i al bi t valu e : 1/4 bi t ,
1/2. bi t , e t c . The se e r r o r s c an also be spe c i f i e d i n t e r m s o f t he le ast si gn i f i c an t bi t ( LSB).
Fo r d i gi t al d i splays, e r r o r s ar e t ypi c ally st at e d i n t e r m s o f le ast si gn i f i c an t d i gi t o r i n pe r -
_____ c e n t o f r e ad i n g.
E R R O R V S . A M B IE N T T E M P E R A T U R E
Pr e c i si o n e r r o r s i n vo lve t he
ki n d o f m at h t hat i s be i n g
d o n e : si n gle pr e c i si o n vs. d o u -
i si o n , i n t e ge r vs. f lo at -
i n g po i n t m at h, e t c .
Ac tu a l E rrors
The e r r o r st at e d i n t he e r r o r
spe c i f i c at i o n s r e p r e se n t s t he
li m i t s o f t he e r r o r , n o t n e c e s-
sari ly t he e r r o r t hat t he d e vi c e
wi ll e xhi bi t i n t he f i e ld . Ac t u -
al e r r o r c an be d e t e r m i n e d by
t e st i n g t he i n st r u m e n t i n u se
an d u n d e r pr o c e ss am bi e n t
c o n d i t i o n s.
Te st i n g a si n gle i n st r u m e n t
d o e s n o t p r o vi d e su f f i c i e n t
d at a t o c har ac t e r i ze a gr o u p o f
t he i n st r u m e n t s, bu t t he r e ar e
st at i st i c al m e t ho d s t hat c an be
u se d whe n a sm all gr o u p o f
i n st r u m e n t s ar e t e st e d t o e st i -
m a t e t he a c c u r a c y o f t he
i n st r u m e n t i n ge n e r al . ' Thi s
c an be p ar t i c u l ar l y u se f u l i n
c o m par i n g m an u f ac t u r e r s.
The ac t u al e r r o r s d e t e r m i n e d
d u r i n g c ali br at i o n c an be c o m -
par e d t o pr i o r c ali br at i o n s f o r
t he sam e i n st r u m e n t t o m e asu r e
t he n e e d t o se r vi c e o r r e plac e t he i n st r u m e n t o r t o c han ge t he c ali br at i o n c yc le . Thi s i s
be c o m in g m u c h m o r e prac tic al wi t h t he ad ve n t o f c ali br at i o n m an age m e n t syst e m s t hat c an
e asi ly r e t ai n t he c ali br at i o n hi st o r y o f an i n st r u m e n t . H
10 20 30 40 50 60 70 80 90 100 110 120 130
ft f.
I
nil
THE ERROR DUE TO AMBIENT TEMPERATURE FOR A GENERIC TRANSMITTER IS A
COMBINATION OF THE REFERENCE ACCURACY AND THE TEMPERATURE ERROR.
DUE TO CANCELLATION, THE TOTAL PROBABLE ERROR IS LESS THAN THE WORST
' ~ AS E ERROR.
R esolution: The smallest interva l tha t c a nbe dis-
tinguished bet weentwo mea su rements. Indig-
ital systems, this is rela ted to the number of bits
of resolutionof ananalog signal, (e.g., the ana-
log spa ndivided by the c ou nt resolution1).
R epeatability: The degree of agreement of a num-
ber of c onsec u tive mea su rements of the out-
pu t for the sa me va lu e of input under the sa me
opera t ing c ondit ions, a pproa c hing fro mt he
same direc tion.1 Note tha t this spec ific a tionis
a pproa c hing fro mone direc tiona nd does not
Include a ny effec t s of hysteresis or dea d band.
T he repea ta bility er r o r spec ific a t io nIs t he
largest error det ermined fr o mboth u pwa rd
a nd downwa rd tra verses. The field or c ommon
usage of this termis closer to the term"repro-
du c ibillt y"tha nt o t he IS A definition.
leprodu c ibillty: T he degr ee o f a gr eement o f
repea ted mea su rements of the ou tpu t for the
sa me va l u e o f inpu t ma de u nder t he s a me
opera ting c ondit ions o ver a perio d o f t ime,
a ppro a c hing fr o mbot h dir ec t io ns . R epr o -
du c lbility inc l u des t he effec t s o f hyst eresis,
dea d band, drift, and repea ta bility.1 This term
Is a nexc el l ent spec ific a tion, bu t for some rea -
son, t he vendo r s ha ve c ho sent ypic a l l y not t o
u se It for mo der ninst ru ment s.
Sampling error: The error c a u sed by sa mpling
a signa l wit h t oo l ow a sa mpl ing frequ enc y. In
genera l, t he sa mpl ing fr equ enc y shou l d be a t
lea st t wic e t he highest frequ enc y int he signa l
being sa mpl ed.
Static pressure effec t error For differentia l pres-
su re t ra nsmit t ers, the perc ent c ha nge inzer o
a nd spa ndu e to a sta tic pressu re change fro m
referenc e c o ndit io ns . T his erro r c a nbe mini-
mized by zeroing the tra nsmitter under actu -
a l opera t ing st a t ic pres s u re.
R E FE R E N C E S:
1. "P roc ess Instru menta tionTerminology," ISA-S51.11993, Instru ment S oc iet y of A mer ic a .
2. Measurement Uncertainty Handbook, O r . R . B . A ber net hy et a l. & J.W. T ho mps o n, I S A 1980.
3. "Is Tha t M ea s u r ement V a l id?," R o ber t F. Ha rt & M a ril ynHa rt , Chemical Processing, O c t o ber 1988.
4. "Wha t T r a ns du c er P er fo r ma nc e S pec s R ea l l y M ea n," R ic ha r d E . T a s ker, Sensors, N o vember 1988.
5. "P erforma nc e Testing a nd A na lysis of Different ia l P ressu re a nd Ga u ge P ressu re Tra nsmitters," Lyie E.
Lofgren, ISA 1986.
6. "C a l ibra t io n: Hea r t o f Fl o wmet er A c c u r a c y," S t eve Ho pe, P.E ., INTECH, A pr il 1994.
Tempera tu re effec t error: T he perc ent c ha nge in
zer o a nd s pa nfo r a na mbient t emper a t u r e
change fromreferenc e conditions. This c a nbe
a signific a nt c ont ribu t or t o t he t ot a l error.
Vibra tioninflu enc e error: T he er r o r c a u s ed by
expo s ing t he inst ru ment t o vibra t ions, no r -
mally spec ified per g of a c c elera tiona nd u p
t o s o me frequ enc y.
JUNE/ I996 CONTROL
A S U R E M E N T
H o,w A c c u r a te
isA c c u r a te ?
P a r t !
f
i.
*
THE ACCURACY OF A MEASUREMENT STARTSWITH THE INSTRUMENT
ERROR SPECIFICATIONS. BUT IN ORDER TO COMPARE INSTRUMENTSOR
SYSTEMS, THE ERRORS FROMDIFFERENT SOURCESMUST BE COMBINED
TO CALCULATE THE OVERALL ACCURACY
By Willia m Most ia
1L Mostia, *, P.L, Is a
senior process control and
computing engineer with
Amoco Corp.'s Worldwide
Engineering & Construction
division In Houston.
This a r t icle is t he second in a t hr ee-pa r t ser ies. In June, P a r t I ex a mined t he concept s a nd t er minology used
t o define a ccur a cy of process mea sur ement s. Here, P a r t II tells how t o combine er r or s wit hin a n inst r ument
or syst em t o pr ov ide a n est ima t e of t he t ot a l er r or . In August , P a r t III will pr ov ide a met hod t o det er mine
how a ccur a t e a n inst r ument or syst em must be t o per for m it s specified funct ion.
O
ne of t he less plea sa nt ex per iences for t he unwa r y equipment specifier is find-
ing out a n ex pensiv e ins t r ument doesn't giv e r ea dings a ccur a t e enough t o
cont r ol t he pr ocess. For t he per son whose signa t ur e is a t t he bot t om of t he r eq-
uisit ion, t he pr ospect of going t o t he boss a nd a sking for a bet t er r epla cement
is, a t best , emba r r a ssing.
In some ca ses, t he ina dequa t e inst r ument st a y s on t he job, giv ing r ea dings no one
r ea lly believes. At wor st , t he ex ist ence of a pr oblem is eit her undiscov er ed or denied,
a nd t he inst r ument becomes a sour ce of t r ouble t ha t ca n r a nge fr om v a r ia t ions in pr od-
uct qua lit y t o unexpla ined equipment ma lfunct ions a nd shut downs.
I n s t r u me n t v e ndor s a r e not a lot of help [see "Fig h t Ag a i n s t 'Sp e c ma n s h ip '
Should Follow Food La bel Lea d," CO NTRO LJune '94, p58]. They t ypica lly a dv er -
t ise t heir inst r ument s in t er ms of r ef- | J MH; HI
er ence a ccur a cy . B ut when y ou look
f u r t h e r , y ou f ind t ha t t he ma nuf a c -
t u r e r h a s p r o v i d e d a mu l t i t u d e o f
er r or specifica t ions t o cover t he r a nge
of pr ocess a nd env ir onment a l condi-
t ions wher e t he inst r ument might be
used. These e r r or s ca n be s ignif ica nt
a nd must be ev a lua t ed befor e t he full
st or y of t he inst r ument 's a ccur a cy ca n
be de t e r mine d.
For exa mple, a 0.05% r efer ence a ccu-
( cont inued on p40)
T W O - C O M P O N E N T S Y S T E M
0-
T M pw a te n tr u M itUr
2S T F
THE ERRO RS O F INDIVIDUAL CO MP O NENTS MUST BE CO MB INED TO GET THE
TO TAL ERRO R O F THE SYSTEM. BUT SIMP LY ADDINGTHE ERRO RS WILL RESULT IN
UNDERESTIMATING THE SYSTEM ACCURACY.
38 J ULY/ 1996
CO NTRO L
(continued f r om p38)
racy over a given r ange m ight be qualified by an am bient tem -
per atur e of 20 2C, at 50% relative hum idity, and within 90
days of calibration. To know the accuracy rating a year later at 28C
and 90% RH, it is necessary to com bine the errors f r om these dif-
f er ent conditions.
So the savvy specifier, given the error specifications or actual
er r or s for an instr um ent or system , needs to know how to com -
bine them for an overall accuracy num ber. This num ber can then
be used to com pare instrum ents, do proration calculations, deter-
m ine sensitivity, analyze system s, and so on.
I n d i v i d u a l I n s t a l m e n t A c c u r a c y
Calculating wor st case er r or (WCE), wher e all the er r or s in
an instr um ent or system ar e added up in the wor st possible
way, gives a lar ge er r or num ber . Field tests on instr um ent sys-
tem s have shown that the actual er r or s are considerably less
than the wor st case er r or s. It also has been deter m ined t hat
these er r or s ar e statistically r andom , whi ch m eans they can
be com bined using the root sum square m ethod (RSS). 2' 3
For exam ple, f or a gener ic dif f er ential tr ansm itter :
Ba s e Con d i t i on s :
Te m pe r a t u r e : 7 5 " F
Static Pressure: 0-psid (pounds per square inch, differential)
Delta P: 100-ln.WC (Inches of water column)
Operating Conditions:
Ambient temperature: 25-125F
Static Pressure: 500-psi
where:
eM = Reference accuracy,
= Combined zero and span shift due to temperature,
= Zero shift due to static pressure,
= Span shift due to static pressure, and
= Other errors as appropriate.
But since individual instr um ent accur acy specif ications ar e
consider ed statistically r andom , they can be com bined using
a r oot sum squar e calculation. Accor ding to such a calcula-
tion, the total pr obable er r or (TPE) f or this dif f er ential pr es-
sure t r ansm i t t er is:
V0.1%2
0.56%
(0.3125% 0.125%)' + 0.3125%2 * 0.125%2
We can see that in diis case, the total probable error is about 5 7%
of die worst case error.
I n s t r u m e n t S y s t e m A c c u r a c y Ca l c u l a t i on s
It also is possible to calculate the total pr obable er r or f or a sys-
tem that contains sever al i nst r um ent s. If no calculations ar e
r equir ed, the RSS m ethod can be used to com bine the er r or s
of the individual instr um ents. You m ust ensur e that all the
er r or s ar e specif ied in the sam e m anner . For exam ple, f or the
system shown in Figur e 1 , the total pr obable system er r or
(TPEsn ) is est i m at ed by:
TPES,
Transmitter Specs:
Calibrated Span = 100-in.WC
Upper Range Limit (URL) = 250-in.WC
Reference accuracy 0.1% of calibrated span (Including the
effects of hysteresis, linearity, and repeatability)
Temperature Zero/Span Shift Error = (0.25% of URL + 0.25% of
calibrated span) per 100F.
Static Pressure Zero Shift Error = 0.25% of URL per JOOO-PSI
Static Pressure Span Shift Error = 0.25% of span per 1000-PSI
The wor st case er r or f or diis dif f er ential pr essur e tr ansm itter is:
WCE (ew + e,Mmn + e^* e,^,...)
= (0.1% *
50 F 250l n .WC
(0.25% x
100'F 100in.WC
= (0.1% + 0.3125% + 0.125% * 0.3125% + 0.125%)
= 0.975%
= J(0.5%' * (0.1% x 52IE. )'
500 F
0.502%
Notice that since the t em per at ur e tr ansm itter er r or is in
per cent of span ( 0-500F) and the display er r or is in per cent
of r eadi ng, the display accur acy m ust be conver ted to a per -
cent of span er r or num ber bef or e it can be used in the e r r or
calculat i on.
If instr um ent er r or s m ust f ir st be calculated, the total pr oba-
ble error calculation is m ore complex because input errors are now
added, subtr acted, m ultiplied, divided, squar e r ooted, etc. The
ef f ects of the these m athem atical oper ations m ust be taken into
account in addition to any inherent er r or in the calculation device.
The calculation er r or can be deter m ined using a Taylor ser ies
expansion. For the gener al case:
D = f(A,B,C)
it can be shown2-3 that the calculation er r or can be estim ated by:
30ej2*(^e,)2*(^ec)2
1 ^A * ' ^D 8 3P C
oA OD do
( cont i nued on p42)
40 JULY/ 1996 CONTROL
( c o n t i n u e d f r o m p40)
to whi c h we m u st ad d the e f f e c t o f die i n he r e n t device e r r o r , e :
Fo r e xam ple , f o r die e qu at i o n :
KxAxB
t he e r r o r i s e st i m at e d by:
This assum es t hat the r e is n o t an y u n c e r t ai n t y associated wi t h
t he "K" f ac t o r .
Calc u latio n s in vo lvin g c o n st an t s t hat have an u n c e r t ai n t y asso -
c i at e d wi t h d i e m , su c h as an o r i f i c e c alc u lat i o n , m u st c o n si d e r
d ie e f f e c t o f t he u n c e r t ai n t y i n t he c alc u lat i o n . In t he Flo w M e a-
su r em en t E ngineer ing Han d bo o k4, die er r or s of die o r if ic e calculation ar e
d iscu ssed at le n gt h. The r e also i s a t r e at m e n t o f se n si t i vi t y c al-
c u lat i o n s, whi c h c an be u se d t o d e t e r m i n e t he e f f e c t o f a gi ve n
e r r o r o r u n c e r t ai n t y i n a c alc u latio n .
In order to m ake die e r r o r calculation f o r a calculating device, die
er r o r s m u st be st at e d in die sam e m an n e r (usually in pe r c e n t o f
here.
http://www.controlmagazine.com
(requires Netscape Navigator 2.x)
span) an d die signals m u st be n o r m alized ( r an ge f r o m 0-1.0). Also
n o t e diat die er r o r analysis m u st consider die o per atin g r an ges o f
all die i n pu t var iables involved so d iat die wor st case var iable val-
u es will be u se d hi die e r r o r c alc u latio n f o r exam ple, die lar ge r
die var iable s o n die to p o f an e qu at i o n an d die sm aller die var i-
ables on die bo t t o m , die lar ge r die calculated e r r o r .
"The Applic atio n of Statistical M e t ho d s in E valu atin g die Ac c u -
r ac y o f An alo g In st r u m e n t s an d Syst e m s," by C.S. Zalkin d an d
F.G. Shi n ske y1, is an e xc e lle n t pape r o n this su bje c t an d pr ovides
e r r o r e st i m at e s f o r a n u m be r o f c alc u lat i o n devices. Som e o f die
m o r e c o m m o n o n e s ar e li st e d i n Table I.
EQUATION
8 = Kx A
C = K,A * K2B
ERROR
B =
= KAB ec = (KBe,)a
-
42 J ULY/ 1996
whe r e e i s t he i n he r e n t d e vic e e r r o r .
Con clusion s
The e st i m at i o n o f an i n st r u m e n t o r syst e m ac c u r ac y c an so m e -
t i m e s appe ar t o be a c o m pli c at e d i ssu e . The e st i m at e o f t he ac c u -
r ac y o f an i n d i vi d u al i n st r u m e n t i s t he c o m bi n at i o n o f t he r e f -
e r e n c e ac c u r ac y an d t he i n f lu e n c e e r r o r s u si n g t he r o o t su m
squ ar e m e t ho d ( RSS).
E r r o r f o r i n st r u m e n t s t hat pe r f o r m c alc u lat i o n s c an be e st i m at -
e d u si n g Taylo r se r i e s e xpan si o n an d RSS m e t ho d s. The e r r o r f o r
an i n st r u m e n t syst e m c an t he n be e st i m at e d by c o m bi n i n g t he
i n d i vi d u al i n st r u m e n t e r r o r s u si n g t he RSS m e t ho d . Usi n g the se
t e c hn i qu e s, yo u c an , with som e r e lative ly si m ple calcu latio n s, esti-
m at e t he ac c u r ac y e n ve lo pe o f an y i n st r u m e n t o r syst e m . H
REFERENCES:
1. "There's more to tran smitter accuracy than the spec," William J. Demor-
est. Jr., Instrument & Control Systems, May 1983.
2. "The Appl ication of Statistical Methods in Eval uatin g the Accuracy of An a-
l og In strumen ts an d Systems," C.S. Zal kin d & F.G. Shin skey, The Foxboro Co.
3. Measurement Uncertainty Handbook, Dr. R.B. Abern ethy, et at. & J.W.
Thompson , ISA, 1980.
4. How Measurement Engineering Handbook, R.W. Miller, The Kin gsport Press, 1983.
5. "Predictin g Flow Rate System Accuracy," Wil l iam S. Buzzard, Fischer &
Porter Techn ical In formation 10E-12, Jan 1980.
6. "What Do Al l Those Tran sducer Specs Real l y Mean ?," Chuck Wright, Personal
Engineering, February 1996.
7. "Performan ce Testin g an d An al ysis of Differen tial Pressure an d Gauge
Pressure Tran smitters," Lyl e E. Lofgren , ISA, 1986.
CONTR OL
Each one of these partial diffentials
should have their associated "e" within
the parenthesis.
"^Accurate 0
isAccurate?
P a r t 3
ACCURATE MEASUREMENTS ARE ESSENTIAL FOR OPTIMUMCONTROL,
BUT EVERY INCREMENT OF ACCURACY COMES AT A COST. HERE'S A
WAY TO DECIDE HOW ACCURATE AN INSTRUMENT TO SPECIFY
This a r t icle is t he t hi r d in a t hr ee-pa r t ser ies. In June, P a r t I exa mined t he concept s a nd t er minology used t o
define a ccur a cy of pr ocess mea sur ement s. In July, P a r t II t old how t o combine er r or s wi t h i n a n inst r ument or
syst em t o pr ovide a n est ima t e of t he t ot a l er r or . Her e, P a r t III pr ovides a met hod t o det er mine how a ccur a t e
a n inst r ument or syst em must be t o per f or m it s specified f unct i on.
H
ow a ccur a t e must a n inst r ument be? The quest ion r eeks of compr omiseint uit ively,
we wa nt t he most a ccur a t e inst r ument we ca n get . But r a ising a ccur a cy t o t he st a t us
of a hoh/ gr a il is t he per oga t ive of philosopher s a nd per ha ps r esea r ch scient ist s, for whom
t he quest ions of cost a nd r elia bilit y a r e piffling t r ifles. Engineer s ha ve t he r esponsibilit y
of choosing die lowest -cost , ea siest t o ma int a in equipment t ha t will get die job done.
The ba sic quest ion is how a ccur a t e a n i nst r ument or syst em ha s t o be t o a void i nf l u -
encing t he mea sur ement or signif ica nt ly degr a ding t he syst em er r or specif ica t ion. In
P a r t II, we sa w how t o a dd er r or s f r om va r ious sour ces t o det er mi ne t he syst em er r or .
For a ny pr oposed i ns t r ument a nd s ys t em, t he er r or s of t he syst em component s ca n be
a dded usi ng t he r oot sum squa r e met hod t o gi ve a n est ima t e of t he syst em a ccur a cy.
But how a ccur a t e does t he syst em ha ve t o be? The a ccur a cy of t he mea sur ing syst em
should be higher t ha n t he r equir ed mea sur ement a ccur a cy. But how much hi gher ?
One wa y t o under st a nd t he impa ct of inst r ument a ccur a cy is t o consider inst r ument
ca libr a t ion. Just a s inst r ument s should be mor e a ccur a t e t ha n t he r equir ed mea sur ement
a ccur a cy, ca libr a t ion equipment must be mor e a ccur a t e t ha n t he inst r ument s t o be ca libr a t ed.
In indust r y, a common r u l e- o f - t h u mb a ccur a cy r equi r ement f or ca li br a t i on is t hr ee
t o f o u r t imes (3 - 4X) mor e a ccur a t e t ha n t he i ns t r u ment per f or ma nce speci f i ca t i on.
Ot her sour ces g i v e ot her r a t i osyou ca n f i nd 2 X' ', 5X\S, 3 - 1 OX4, 4- l OX! , o r
ot her guidelines. AP I MP MS CHAP TER 2 1 - "Flow Mea sur ement Using Elect r onic
Met er i ng Syst ems, " f or exa mple, st a t es t h a t t he ca li br a t i on st a nda r d should be t wo
t imes bet t er t ha n t he mea sur ement speci f i ca t i on5. ISA St a nda r d SS1 . 1 , on t he ot her
ha nd, st a t es t ha t f or gr ea t er t ha n 1 OX, t he er r or of t he ca li br a t i on device does not ha ve
t o be consider ed; bet ween 3 X a nd 10X it must be consider ed; a nd if less t ha n 3 X, t he
ca libr a t ion device should not be used.
Wher e do t hese number s come f r o m? The r oot sum squa r e (RSS) equa t i on, whi ch
(cont i nued on p68)
B y Wi lli a m Mos tia
CONT ROL
William L Mtttia. Jr., P.L, Is a
senior process control and
computing engineer with
Amoco Corp.'s Worldwide
Engineering & Construction
division In Houston.
AUGUST / 1 9 9 6 6 7
( c o n t i n u e d f r o m p 6 7 )
t ake s t he squ ar e r o o t o f t he su m o f t he squ ar e s o f t he e r r o r s, m e asu r i n g syst e m :
sho w; t he r e aso n i n g:
TPE= (-)' e 2
whe r e : whe r e : e = t he e r r o r c o n t r i bu t e d by t he m e asu r i n g i n st r u
TPE = t o t al pr o bable e r r o r m e n t o r syste m .
tt = t he e rro r c o n tribu te d by t he m e asu rin g i n st r u m e n t o r system . X = t he r at i o o f t he e r r o r o f t he m e asu r i n g i n st r u m e n t o r sys-
e = t he e r r o r c o n t r i bu t e d by t he c ali br at i o n d evice. . t e r n t o t he c ali br at i o n d e vic e e r r o r .
If we r e wr i t e t he c ali br at i o n d e vi c e e r r o r i n t e r m s o f t he No w le t 's d e f i n e t he r at i o o f t he e r r o r e f f e c t , R t o be :
, is the complete solution forcafibration
r '
.
and associated datarmaintgriaae Including 'portable and
workshop/laboratory calibrators as well as DOS and Win-
of the calibration process including scheduling, data logging
of calibration results, and generation of various calibration
arid history certificates/reports. !_---- \
TPE-e
Substituting:
R. =
R, = xvx2 * 1 + x!
Fo r e xam ple , c o n si d e r an i n st r u m e n t c al-
i br at e d f o r 0-100-i n .WC t hat has an ac c u -
r ac y o f 0.1% o f span . Usi n g a c ali br at o r
wi d i ac c u r ac y 0.04% o f r e ad i n g:
TPE= yOU)1 * (0.04x >
TPE =v0.0118 = 0.1077%
The e r r o r c o n t r i bu t i o n o f t he c ali br at o r i s
0.007 7 %. The e r r o r e f f e c t r at i o i s:
e 0.1
R. =
TPE-e
&
0.1077-0.1
12.98
R. = 2.5,/(2.5)2 1 * (2.5)2 =12.98
[ The R_ f o r so m e c o m m o n valu e s o f X ar e
sho wn i n Table I :
Re
2.5
8.5
18.5
X
4
5
7
10
Re
32.5
50.5
98.5
200.5
Fr o m t hi s o n e c an se e t hat f o r an "X" valu e
sli ght ly gr e at e r t han 2X ( 2.182, t o be e xac t ),
t he e f f e c t o n t he base ac c u r ac y i s 10:1 o r an
o r d e r o f m agn i t u d e . Fo r 3 -4X t he e f f e c t i s
abo u t 18:1 t o 32:1. an d f o r abo u t 7X ( 7.053)
t he e f f e c t i s 100 o r t wo o r d e r s o f m agn i t u d e .
Ext e n d i n g t hi s t o a m o r e ge n e r al ac c u r ac y
CO NT RO L
FOR FREE INFO CIRCLE 107 ON CA RD
**w
M '
requirement, die minimum "X" is a litde over 2, 3-4X should be ade-
quate for most things, SX is appropriate for higher accuracy require-
ments, and 7-1 OX or more should be used for very high accuracy.
However, die practical use of this information for process field
instruments calibration may be somewhat limited. Modern field
electronic process instrumentation reference accuracy is typically
in die 0.1% range widi transmitters coming on die market in die
0.03-0.08% accuracy range. Many of die field calibrators on the
market have accuracy ratings in die 0.02-0.08% of reading range
and deadweight testers are in the range of 0.01 -0.05% of read-
ing. From diis information, it is not hard to see diat getting past 3X
for some instruments might be radier difficult and cosdy.
In evaluating the selection of instruments for measurement
and control functions based on accuracy considerations, we must
determine what we consider to be a significant effect on the
overall system. For example, if a control specification is 0.2F
and we choose an instrument whose estimated error contribu-
tion is an order of magnitude less than the spec (Rr= 2.2), then
we would expect the probable error contribution by the instru-
ment to be 0.02F. If diis is not considered a significant contri-
bution, dien die instrument's accuracy is acceptable.
We must also remember diat we are typically talking about sys-
tems that contain a number of components, each of which can
contribute to die overall system error. Component errors must be '
added as described in Part II. The system accuracy must also be
evaluated for die expected operating range as outlined in Part I.
Then Table I can be used to help select instruments that will keep
die system accuracy within die engineering requirements. 3
A C C U R A T E REFERENCES:
1. API MPMS 21.1.8.6, "Calibration and Verification Equipment." American
Petroleum Institute
2. "Standards, Simulators, and Calibrators," Measurement & Control, Septem-
ber 1992
3. "Achieving Provable Accuracy in Deadweight Pressure Testers," R.C. Evans,
Measurement & Control, December 1986
4. "Process Instrumentation Terminology," ISA-S51.1,1993
5. "High Accuracy Differential Pressure Calibration," J. Anthony Comeaux,
ISA 1986
6. "There's More to Transmitter Accuracy Than the Spec," William J. Demor-
est, Jr., Instrument & Control Systems, May 1983
7. "The Application of Statistical Methods in Evaluating the Accuracy of
Analog Instruments and Systems," C.S. Zalkind & F.G. Shinskey, The
Foxboro Co.
8. Measurement Uncertainty Handbook, Dr. R.B. Abernethy et at. & J.W.
Thompson, ISA, 1980
9. "Predicting Flow Rate System Accuracy," William S. Buzzard, Fisher &
Porter Technical Information 10E-12, January 1980
10. "What Transducer Performance Specs Really Mean," Richard E. Tasker,
Sensors, November 1988
11. "Calibration: Heart of Flowmeter Accuracy," Steve Hope, P.E., InTech,
April 1994
CO NTRO L
It's time you went to moss!
Need a good reason to
go to mass? Now there's
BASIS from Micro Motion.
A new line of flowmeters
that makes the benefits of
Coriolis technology even
more affordable.
Need more good reasons?
Fast, simple installation
Reduced maintenance costs
High accuracy and
repeatability
Broad range of applications
Award-winning service
and quality
-\ FRO/V;
Time for more
information? Call
l-800-760-81ia
5 Outside of the U.S.A.
303-530-8400.
\o Motion
FO R FREE INFO CIRCLE 133 O N C A R D

Anda mungkin juga menyukai