Вы находитесь на странице: 1из 26

¢ ,$

F
' '
ERROR 5¢,,q
PROPAGATION EQUATIONS FOR ESTIMATING a 2 o 8" O.IN
THE UNCERTAINTY
HIGH-SPEED WIND 'FUNNEL TEST RESULTS*

Edward L. Clark ¢
Sandia National Laboratories

Albuquerque, New Mexico _ (._ _.._._ _)_ / _)_ ..../

.Abstract M Mach number, Via


M aerodynamic pitching moment
Error propagation equations, based on the Taylor se-
ries model, are derived for the nondimensional ratios and Mq derivative
pitch rate, of pitching moment with respect to
OM/Oq
coefficients most often encountered in high-speed wind p pressure
tunnel testing. These include pressure ratio and coefficient, p vehicle roll rate
static force and moment coefficients, dynamic stability co- P pressure ratio, p/Pr _
efficients, and calibration Mach number. The error equa- q dynamic pressure, pVZ/2
tions contain partial derivatives, denoted as sensitivity q vehicle pitch rate
coefficients, which define the influence of free-stream R aerodynamic ratio
Mach number, M**,on various aerodynamic ratios. To fa- R calculated test result
cilitate use of the error equations, sensitivity coefficients RA stream-tube area ratio, Ao./A*
are derived and evaluated for five fundamental aerody- P,g gas constant,
reference areacp-Cv
namic ratios which relate free-stream test conditions to a T absolute temperature
reference condition. V flow velocity
xi ith parameter in test result

Nomenclature _, ratio of specific heats, Cp/Cv


Axi interval for numerical e_,,aluation of sensitivity
Primary symbols coefficient
0 absolute sensitivity coefficient, Eq. (5)
aA speed of sound area of stream tube or channel
cross-sectional 0" relative sensitivity coefficient, Eq. (8)

Cp specific heat at constant pressure p mass density


Cv specific heat at constant volume pxix j coefficient of correlation
CF force coefficient, Eq. (71)
Ct rolling moment coefficient, Eq. (85) Subscriots
rolling moment dynamic stability coefficient,
CtP Eq. (85) a atmospheric
g gage pressure measurement
Cm pitching moment coefficient, Eq. (74) r reference condition
p_tching moment dynamic stability coefficient, t total (stagnation) condition
Cmq Eq. (76) 1 conditions just upstream of a normal shock wave
Cp pressure coefficient, (P-P.o)/q** 2 conditions just downstream of a normal shock
reference diameter wave
E error, either bias or precision oo free-stream condition
E' error from single source
f function Sul)erscriDts
F
l aerodynamic force
reference length * critical conditions (local speed equal to local
L aerodynamic rolling moment speed of sound, i.e., M= 1.0)
Lp derivative of rolling moment with respect to roll
,'ate, OL/3p Introduction

Evaluation of the uncertainty in an experimental result


has always been an important, but frequently neglected,
*This work was supported by the United States Department of
Energy under Contract DE-AC04-94AL85000. part of the test process. With recent emphasis on quality
and traceability of results, it has become essential. Some
tSenior Member of Technical Staff, Associate Fellow AIAA
journals, such as the Journal of Fluids Engineering pub-
This paper is declared a work of the U.S. Go_,ernmet,*and is
lished by the American Society of Mechanical Engineers
not subject to copyright protection in the United States.
(ASME), will no longer accept experimental results unless

MA .....
ooou.....- |1 " Ip'l ' II ' '
they include uncertainty estimates. The experimentalist the final result. The Taylor series method requires that the
who has learned to use uncertainty estimates finds them in- test result functional relationship be differentiated, with re-
valuable both before and "after testing. Pre-test estimates spect to each of its parameters, xi, to define a set of sensi-
are very useful in selecting the instrumentation and test tivity
• coefficients, 0 xi . For most experimental
. aerodynam-
techniques which will best accomplish the test goals. Post- _cresults, the calculus and algebra requtred to evaluate the
test estimates provide the user with error bounds which can sensitivity coefficients are difficult. This is especially true
be indispensable if the data are to be used in design studies, for the ser_sitivity coefficient, 0M.*, of the free-stream
Several organizations including the ASME, Insmtment So- Mach numl:er, M**.In this paper, error propagation equa-
ciety of America (ISA), and the International Organization tions are derived for many of the nondimensional ratios and
for Standardization (ISO)have established standards which coefficients used in high-speed wind tunnel testing. Most
define the uncertainty estimation process. Aerospace orga- of the results are also applicable to low speed wind tunnel
nizations will soon be joining this group as the Advisory tests, but emphasis is on high-speed testing where Mach
Group for Aerospace Research & Development (AGARD) number is an important parameter. Equations and plotted
and the American Institute of Aeronautics and Astronautics values of the sensitivity coefficients, 0 M , are provided as
(AIAA) develop their standards for uncertainty analysis, functions of M_. An extended coverage _f this subject, in-
cluding additional aerodynamic ratios and Reynolds num-
Procedures for estimating the uncertainty in experi- ber, with tabulated results is presented in Ref. 7.
mental measurements are well documented t'6. Although
there are some differences in the methodologies recom-
mended by various authors, it is generally agreed that the ErrorPropagation Eo_uations
procedure can be divided into three steps: (1) estimate mea-
surement errors, (2) propagate measurement errors to final T_ylgr series model
results and (3) calculate uncertainty. In the f'wst step, the
experimentalist defines all independent measurements in- The previously cited references 1-6 all recommend that
volved in the experiment. The elemental errors associated the Taylor series model be used for error propagation into
with each measurement are then defined, according to cat- a result. Detailed derivations of the model are provided in
egories such as (a) calibration, (b) data acquisition and (c) Refs. 1, 2, 4 and 5 and will not be repeated here. Consider
data reduction. At this point the errors must be classified as a test result, R, which is a function of n parameters, xi,
bias (fixed or systematic) or precision (random) errors. Fi-
nally, the elemental errors are combined by root-sum- R = R (x t, x2, ..., xn) (1)

squaring, to provide an estimate of the total bias and preci- Then, the bias or precision error, ER, in R is approximated
sion error for each measurement. In most cases, the de- by
sired test result is not the actual measurement, but is a cal-

2 OR E 2 ...+ _)R E
example,result
culated a nondimensional
incorporating one
ratioor or
more
coefficient.
measurements,
Thus, the
for ER
2 = [,_DR E xlj + (,_ x2) + (,_ xn)
second step in the uncertainty analysis is to estimate the er-
ror in the test result using an error propagation method +2 _R 3R
based on a Taylor series model. The bias and precision er- _XI_X2 Px_x2Ex3Ex2+ "" (2)

rors are propagated separately. Finally, in the third step, the where E x is the bias or precision error in xi. As was men-
bias and precision errors obtained in the previous step are tioned in ihe previous section, bias and precision errors are
combined to provide an uncertainty estimate for the test re- propagated separately and are then combined to estimate
suit. The two factors to be combined are the bias error and the uncertainty in the result. The cross product terms pro-
the precision error multiplied by t, the 95th percentile point vide the error contribution due to correlated errors and
tor the two-tailed Student t distribution. The combined un- Px x is the coefficient of correlation between x 1 and x2.
certainty is estimated by direct addition of the two factors, i 2 . of p .is difficult,
Evaluation . and Coleman and Steele 8 have
or by a root-sum-squared addition, suggested an alternate equation,

The purpose of the present work is to facilitate, for


high-spe_ aerodynamic testing, the calculations required
in step two, that is, propagation of measurement errors to

2
P q
i i

or subtracted, e.g., R=(Xl+X2)/x3, the relative error form is

ER2= t_I(0RExl ./2 +t_-_2(0R


Ex2)']2+... + t_-_n(bR
Ex°)_2 fined,asless
useful. The relative sensitivity coefficient, 0'i, is de-

20RORE, -' , xi0R


+ _-xl-_2 xl_ _2+ "'" (3) 0 i = R 0xi (8)
where E' x.i and E'x 2 ,are the portions of the error ill x ! and Then,
x2 that arise from tlie s,-une source and are presumed to be

ply than Eq. (2). If the correlation between parameters is a


Peffectlyc°rrelated(P=l'0)'Thisequati°niseasiert°ap" (_)2 = ( 0'i Ex_/2
x I } E-_x2)2+
/ 0' 2 xz ) +"" + / 0" E_I 2 (9)
result of the measurements, precision errors are assumed to n xn
be independent (p=0). However, if the correlation occurs as hi many cases, analytical evaluation of the sensitivity coef-
the result of multiple occurrences of a parameter within the ficients is difficult or hnpossible. Examples include results
equation for the test result, even precision errors will be which involve: (1) complex and/or implicit functions; (2)
correlated and Eq. (2) or (3) must be used. In any case, if interpolation of tabulated functions and (3) iterative calcu-
the bias errors are not independent, Eq. (2) or (3) must be iations. The procedure in this case is to use a numerical
used. In evaluating the partial derivatives, it is easy to ig- evaluation of the partial derivatives. Using a central differ-
nore the sign of the derivatives since most of the terms are encing scheme, the derivatives can be approximated by,
squared. However, it is essential that the signs be retained
for the correlation terms which are not squared. OR = lim " (10)
0x i Axi-) o [Rx, + ax,12-Ax Rx,
i - Ax,121
..lXj= const (j _i)
If the errors are independent, Eq. (2) can be simplified
to, With this method, nominal values are selected for all of the
parameters, xi. Then, R is sequentially evaluated by per-
:Z r3R E )2 /3RE )2 r_)R E _2 turbingeachxi(i=l,n)by+Axi/2while_e_maini.gn. 1
ER = _,_11 xl) +_3"x'22 x2) +'"+[_3--_n x,) (4) xj values remain constant. The two perturbed values of R
are differenced and divided by zL_i. The approximate deriv-
The partial derivatives are denoted as absolute sensitiv- atives are used with Eq. (4) to estimate the error in the re-
ity coefficients, 0i, that is, suit. This approach is di_ussed in detail in Refs. 9 and 10.
OR The advantages of this method are that it avoids complex
0i = _ (5) differentiation and 'algebra, automatically calculates the ef-
t fect of correlated variables, and may be the only solution
Then, possible. The disadvantage is that it provides only a"point"
solution and the functional variation of test result error with

ER 01Exl .. ()2 (6)


2 = ()() 2 + 02Ex 2 2+. + 0nEx" apowerful
given variable
and wascannot
used be
in determined.
the present The
reportmethod is very
to verify the
Dividing each side of Eq. (4) by R 2 gives the "relative er- relative sensitivity coefficients for each of the applications.
ror" form of the propagation equation, For these checks, the increment was Ax i = 0.005 xi and the
numerical results were in excellent agreement (4 to 5 sig-

(ER) 2 = [X,3R
R3x I Ex_]2
xI + (x.223R
R 3x 2E_x2]2
x2 } nificantdigits) wilh this
which demonstrates the technique
analytical isvalues. A Ref.
listed in simple
7. code

+ +(X,3R Ex,/2
77,, -
Aoolicatio,i of the ,nodcl
Applicationof the Taylor series model is straightfor-
This form of Ihe equation is especially useful if the mea- ward. However, there _tre two cautions. First, the correct
surement errors arc defined as relative errors, Ex/x i, and if model must bc used. If the measurements are independent,
the test result p:u_uneters :u'e related only by muliiplication Eq. (4) c_m be used, but if they are not independent, i.e., if
_md division, e.g., R=x i x2/xa. When parameters are added two or more of the mt_tsure_aents are correlated, then either
Eq. (2) or (3) ii|ilSl bc used. This is very imlx_rlant eald will
be emphasized again in later sections. Second, it is essen- Example - pressure ratio
tial that the data reduction equation t_esolved for the exper-
imental result before performing an uncertainty aqalysis II. Independent pressure measurements
This mistake is most likely to occur during a pre-test anal-
ysis to estimate the allowable error in a measurement for a As an example of the use of error propagation equa-
given error in the result. For example, let tions, consider the simple pressure ratio, P,

R = xy (11) P = P (15)
Estimate the allowable error, Ex, in the measurement, x, for Pr
a given error, ER, in the result, R. From Eq. (4), where p and Pr are two independent absolute pressure
measurements. Typically, in a high-speed wind tunnel test,
(i)R E _2 (/)R E _2 the reference pressure, Pr, could be p** (measured on the
E2 = \b-x x/ + \_-_ y} sidewall or with a probe), Pt or • Pt 2 •The measurements , are

= (YEx) 2 + (xEy) 2 assumed to be independeat - this may not be the case ffthe
same standard was used to calibrate both transducers
= R2 +
[(._)E 2 (_)21 and/oriftheysharecommonsignalconditioning.FromEq.
(4),

Then, [(/)P "12 (/)P E /21 1/2


Ep = [_,_Ep7 +[fi-p-_p,).1 (16)

Ex = x R J - (12) Differentiating Eq. (15) and substituting the results in Eq.


(16) gives
This is the correct estimate of Ex. However, if the original

resulting equation, the result will be error. Solving Eq, .,Ep p


equati°n f°r R is s°lved f°r x and Ex is
thein
estimated fr°m Ep = [(lp_. )2 + ( r1211/2_
p_E
(11) for x gives,

R and simplifying,
x = - (13)

The error in x is given by, y Ep = _r[E2+


1 ( PEp, )21/2
] (17)

E2 = /)X
_R,).-. ")2 +_._--_Ey)
(_X "_2 ha the relative error form,

= + (18)
= + (SE,;
. P L_" kPr; J
This is the simplest form of error propagation and shows
Then, that the relative error in the result is the root-sum-square of
the relative errors in the two measurements.

Ex = x + (571
y/_l (14)
Correlated pressures
Comparison of Eqs. (12) and (14) shows that the second re-
sult is incorrect. Now, assume that the two pressure measurements are
gage, not absolute. Then, the two absolute pressures re-
quired for the ratio are given by

P = Pg+Pa
(19)
Pr = Pr, g + Pa
i

where pg and Pr,g am the two independent gage pressure (" "_2 1/2

measurements andthePacalculated
sure. In this case, is the measured
pressuresatmospheric
p and Pr arepres-
cor- Ep = El' + (,_r P') J

related by Pa. The pressure ratio, P, is given by, Then, differentiating Eq. (15) ,and simplifying,

Pr, g + Pa Ep = P + (25)
kPrJJ
Calculation of the error in P is much more complicated than
in the previous case, where p and Pr were independent. The The effect of errors in Pa is concealed in this equation. Ex-
error will be evaluated analytically, using two techniques, panding p and Pr, Eq. (19), and estimating their errors gives,

and numerically. E2 = E2 + E 2 (26)


P Pg Pa

Analytical evaluation - indeoendent variables and

If the required calculus and algebra associated with de- E2 = E2 + E2 (27)


termining the sensitivity coefficients permits, it is always P' P'._ P"
best to express the result in terms of the individual mea- Substituting these expressions into Eq. (25) gives
surements, that is, in its most elementary form. This avoids

is demonstrated
the problems in the
which can next
occur section. Since pg, variables
with correlated Pr,g, and Pa
as Ep = P [(_)2 + k(Ep_._)2
Pr) + (1_ + _r2) Ep2,]112 (28)
are independent measurements, Eq. (20) is the desired form
and the error estimate is given by Eq. (4), Comparing Eq. (28) to Eq. (23) shows that there is a miss-

F( II'_P E I'_9 |or('" E II_9"_


,/2 ing term, -2E-2v,/PP,"Therefore, Eq. (28)wi, overpredict
Ep= Lt__P Ep,j +_,_--_r,g P"') +_,_P, P')I (21) the erroras in
treated Ep. Thereason
independent forthisis
variables, whereasthatthey
p and Pr were
are actually
correlated by Pa.
Performing the indicated partial differentiation and simpli-
fying gives, For a correct error estimate with correlated variables,
use Eq. (2) or (3). Using Eq. (3),

EI,=P
_,W) \P-Or ) PiJ
P tO_Pr Pr)
or,
+ 2 °-}p°3p E' E'
l' p(Pa) (29)

Ep = P (Ep, _
+ _
1 1
+ _ + Pr2
2 "_E2
PPr) p' i
(23) where E'p and E'p are the portions of the error in p and Pr
that originate fro,d the correlating factor, Pa. From Eqs.
(26) and (27),
Notice that in this equation, the pressures in the denomina-
tors are the absolute pressures p and Pr, not the gage pres- E' =
sures pg and Pr,g. Therefore, this is not the usual relative er- p Ep,
mr form. E'pr = Ep,
The sensitivity coefficients are,
An_dvtical evaluation - deoendenl v_u'iables
_P_ 1 and _)P - P
Frequently, it is easier to differentiate the unexpanded OP Pr ()Pr p2
equation, Eq. (15) in this case. If we were to take this easier
course and treat p and Pr ,'ts independent wuiables, we Substituting these relations into Eq. (29) and simplifying
would obtain the error in P ;ts, gives,
!
, t

EI, = P + _ + + (30) the absolute error


Or J _ 02 PPrJ P'J

OR _ "]2 2 (35)
which agrees with Eq. (23). • Ep. = _M_M_

Numerical evaluation The sensitivity coefficient for lVlach number is,


Dp_ OR
A program utilizing Eq. (10) was used with Eq. (20) to 0M.. - _ = Pt DM** (36)
estimate the three sensitivity coefficients - DP/bpg, DP/0Pr,g
and DP/Dpa.The numerical results agreed with Eq. (23). Dividing both sides of Eq. (35) by p**gives the relative er-
Ix)r,

Sensitivity Coefficients - Aerodynamic Ratios


The aerodynamic ratios considered in this section are
basic to high-speed aerodynamics and, in general, relate
Ep**
P--_= [( M DR E M
"if" DM**M**) + : Pt ) _!
(37)

calculated free-stream conditions to measured total (stag- The relative sensitivity coefficient for _ is
nation) conditions. There are two reasons for interest in M** DR
these ratios. First, they provide a means of estimating the 0'M** = R DM** (38)
uncertainty in the calculated bee-stream conditions. Sec-
ond, they are required for evaluating the uncertainty in the In evaluating Eq. (35) or (37), if Pt is calculated from a gage
more complex applications which are discussed later. In all pressure measurement, Pt,g, then

cases, these aerodynamic ratios are functions of only the Pt = Pt, g + Pa (39)
free-stream Math number, M_, and the ratio of specific
heats, y. In the previous section, the uncertainty in the pres- where Pa is measured atmospheric pressure. The error in Pt
is
sure ratio, R = P/Pr, was examined. In that case, both pres-

total condition is measured-the free-stream value is to be E = E + E


sures were from
calculated measured values. In theratio,
the aerodynamic present section,
R, and only con-
the total the Pt L_.DPt, g Pt.g )2 /0pt
_.Dpa P,J)21J 1/2
dition. For example, consider the ratio of free-stream pres-

sure, p._, tO total pressure, Pt, Ept = [E2t, g + E2] 1/2 (40)
P..
R - - f (M._,)9 (31) This value of Ep, would then be substituted in Eq. (35) or
Pt (37).
Then, the free-stream static pressure is calculated with The analysis in this section makes several assumptions.
p** = R (M**, y) Pt (32) First, it will be assumed that Pt is an absolute pressure mea-
Assuming that the variables are independent, the error in surement (or equivalently, that Ep is negligible). If this is
not the case, Eq. (40) can be us_ with Eq. (35) or (37).
p, is given by Second, it will be assumed that there is negligible error in

= E (33) dependent. This may not be true in many tunnels where the
Ep.. [(Dp__
0-_EM..J "_2+ [Dpo. ")2+ _.Dpt
_-_ E_,) )211/2
(0p. * PtJ J 7' i'e" E_'= 0"
free-stream Finally'
Mach it is isassumed
number that from
determined M_' and pt are in"
a calibration
relation which is a function of total pressure. In this case
Letting F_
3,= 0 and expanding the partial derivative in the
there are three options. First, expand the equation of inter-
first term (chain rule), est to its elemental parameters, which are independent, and

Ep.. = [(0p_/)R
DR DM EM- )2 + _tt
(Dp_)2]1/2
Ept (34) differentiate
the sensitivity
to define it coefficients.
the errorwith respect Equation
equation. to each modify
Second, parameter
(4) canthethen
to evaluate
be equa-
error used
tion to include the effect of correlated parameters, e.g., use
i

Eq. (3). Then use the sensitivity coefficients for Mach


number, which are given in the following sections, with the 0R - T M..R (42)
appropriate error estimates defined by the M**calibration 0M,,, 1 + _21M 2
relation. Third, use a numerical evaluation which defines
sensitivity coefficients that include the effects of correlated
The relative sensitivity coefficient is
pm'ameters. These coefficients can be used with an error

equation based on Eq. (4). M _)R T M2


0' = - (43)
In tile following subsections, the derivative 0R/OM.,, M. R 3M** 1 +'__ IM2
2 ,,_
_md the relative sensitivity coefficient, 0'M=
_)(3R/gM**) will be evaluated as functions of Moofor For 7 = 1.40, values of R, OR/DM_ and (M**/R) OR/OM**)
five fundmnental aerodymunic ratios. Values of R, m'e plotted in Figure 1.
3R/DM**,and 0'M, are plotted fo" y = 1.40. The calculated
values of R were checked by comparing them to the v_dues
given in Ref. 12. Analytical values of 3RfdM,_ were Free-stream dynmnic pressure ratio, R = q**/ Pt
checked by numerical differentiation of the tabulated val-
The free-stream dynamic pressure, q**, is calculated
ues of R. from

Eree-streamstatic pressure ratio,R = p**[ Pt q** = R (M,_, T) Pt (44)


and the error equations are
The equations for free-stream static pressure, p**, and

its absolute and relative errors are given above by Fqs. OR .-, ')2 2 (45)
(32), (35) and (37), respectively: Eq,., = Pt _-'_M_ + REp,

p,_ = R (M,_, T) Pt [32] and

[( 0R _ _2 ]1/2 --= + (46)


For an isentropic flow of a perfect gas, Eq. (48) of Ref, 12
gives,

= 2M** l+-- (47)

Differentiating Eq. (47) with respect it) M,_ gives


For an isentropic flow of a perfect g_tsI , Eq.(44) of Ref. 12

DM - _"T'----iL2 / (48)
gives, 0R M_( 2-M2_ )
R - P-_ - 1+ - M_ 7-1 (.41)
Pt
The relative sensitivity coefficient is
Differentiating Eq. (41) with respect to M_ gives
M OR 2-M2_
= - (49)
0'M
IFollowing the notation of Ref. 12, a "thenmdly R DM 1 + T---_IM2
perfect" gas is one which obeys the thermal equa- 2
tion of state, p = pRgT. A "c_dorically perfect" gas For T = 1.40, values of R, 3R / 0M_ and (M_/R) (3R/3M_)
is one which has constant specific heats, Cp and are plotted in Figure 2.
Cv. A "perfect" gas is both thermally and calori-
c_dly perfect.
r

, s

Pilot pressure ratio, R = pt2/ph Ravleigh pitot ratio, R = p,ffPt2

The pitot pressure, Ph' is calculated from The Rayleigh pitot ratio is normally used, not to calcu-
late p**or Pt.' but to calculate _ from the two measured
pt 2 = R (Moo, T) Ph (50) pressures. Tl_is application will be discussed in more detail
The error equations are in a later section. At free-stream Mach numbers less than
1.0, Pt = Pt and R = P.o/Pt. For M_> 1.0 with adiabatic

[( OR.-. "_2 ]1/2 flow o' a perfect gas, Eq. (100) of Ref. 12 gives,
_L

and R= P*°
[--2 = Pt2
Pt Pl =
-- (T + 21) M 21'-1

Pt---_= I/ --if- 3M*oM_ ell


+ _,_ (52) x L T+l (56) I

For free-stream Mach numbers less than 1.0, Pt = Pt Differentiating F--xt.(56) with respect to M,,. gives
> 1.0 with
. . .
and R= 1.0. For M**_ adlabauc flow of a 2perfect
1

gas, Eq(99)ofRef. 12gives, __.L. OR


DM_ M_L2TM_
R r 2T(1-2M2)]
(T- 1) (57)

,:p,:i/,+l,*
],
Pt1 (3'- 1) M_ + 2 The relative sensitivity coefficient is

+ 1 -t-I 0' M** DR _ -

For T = 1.40, values of R, DR/DIVI_and (M,,./R) (DR/3M,,.)


Differentiating Eq. (53) with respect to M_ gives are plotted in Figure 4.

DR _ 4TR 1- 1 Stream-tube area ratio, R = A=,/ A*


DM Y-1LM [(T_I)M_+2]
The sU'eam-tube area ratio will be used in the applica-

- 27M_- M (T- I) _1 (54) lions to estimate


ofaperfect the error
gas, Eq. in Ref.
(80) of q_ / V_. For an isentropic flow
12 gives

The relative sensitivity coefficient is R = _, = _ 1+ M2 )]'+'


2 (T- I) (59)

0'M = MR 3RoM 4TM_T


-1 ,[Moo[ (T- 1)
2]1M2 + Differentiating Eq. (59) with respect to _ gives

M** DR _ R oo (60)
- (55) 3M -
2TM2 - (T-- 1) "' M 1+ M

ForT= 1.40, values of R, DR/DM_ and (M JR) (3R/OM,_)


The relative sensitivity coefficient is
are plotted in Figure 3.

Moo0R M 2,_- 1
= - (61)
O'M.. R OMoo 1 +Y-IM2
Too
I •I •

i
+

For y = 1.40, values of R, OR/0M** and (M.+/R) (DR/01VI**) Pt, using Eq. (41). Assuming that the variables in Eq. (63)
are plotted in Figure 5. are independent, the error in P is given by
The stream-tube area ratio can also be used to estimate

sectional
the sensitivity
area and
of Mo,
to changes
to changes
in the
in the
test sonic
sectionthroat
cross-sec-
cross- Ep = P [(_ + / E__p.p_
PooJ) 2 J] 1/2 (64)
tion,'d area (e.g., blockage at subsonic speeds). Since

R = f (Moo, y) where Ep is the error in the measured pressure and Ep+. was
given by Eq. (37) as,
DR
ER
Ep+. DR EM ]2 Ep t
Except for _ = 1.0, where 0R/0M..= 0, P"-Z= _M**M ") + k._ [37]

r DR __l where R = p.,/Pt, EM is the error in the free-stream Mach


/ / number, and Ep is th_error in the measured total pressure.
EM,* = _j ER (62) The relatwe
. sensttlvlty
, ..... coefficient, (M.o/R)(0R/0M**), ts
given by Eq. (43) and is plotted in Figure 1 for y=1.40.
Then, with 0R/0M** from Eq. (60), the sensitivity of M,o to
stream-tube area ratio, R - A,,,/A*, can be estimated with
Eq. (62). Pressure coefficient, Cp

The pressure coefficient, Cp, is defined as

6mlJ.d.0_ p-p.+
Cp = (65)
In this section, error estimates will be given for several q**
nondimensional ratios and coefficients which are frequent- For high-speed wind tunnels, the dynamic pressure is usu-
ly used in presenting the results of high-speed wind tunnel ally calculated from the measured total pressure and the
tests. These estimates will rely heavily on the results of the flee-stream Math number with Eq. (47). The static pres-
previous section. It will be assumed that the variables are
sure, p**is calculated _'ith Eq. (41). For a thermally perfect
independent. For those results which involve both an arbi- gas, q**and p**are related by Eq. (47) of Ref. 12,
trary measured pressure and a measured tunnel condition
_'_ M E
pressure, such as total pressure, Pt, this implies that the two q** = _r'.o ** (66)
pressures are absolute measurements or are gage measure-
ments with a negligible error in the measured atmospheric Then, substituting Eq. (66) in Eq. (65),
pressure. If this is not the case, a more complicated analysis
will be required. Also it is assumed that Mo. and Pt are not C = p - p+*
correlated through the tunnel calibration. If this is not true, P Y.. M 2
the equations may be much more complex and a numerical _'oo oo

estimate
is no errormay be ratio
in the required. Finally,heats,
of specific it is assumed
y. that there = yM_
_.__2( _p - 1]

_Pressure ratio, P and substituting p**= R(M.o,y) Pt gives

For the first application we wiU again examine that Co = _M-_(pt_- 1) (67)
ubiquitous parameter, the pressure ratio, P, where

Assuming p, Pt and Mooare independent, the error in Co is


P = -P-- (63) given by
Poo

ln this equation,
is c',dculated fromp is an arbitrary
M,,,, y, and themeasured
measured pressure and p+o
total pressure, Eco = [(i_2p
_ E p)2 + /_
_,0PtP Ep,))2 + /OCp )2 ](68)j
_.0Moo EMoo) 1/2
k
t

( t/20j
p= Cm
i/ /2
• ( 112"2'75
R _--_.] 72/ /2]
_, M_, ) (69) This equation
measured is similar
moment to Eq.
and Et (72) with
the error in theE M the errorlength.
reference in the

where P = p / P.o and R = pc,/Pt. If desired, P can be re- Frequently, l -- d and S = _i2/4. Then, I and S are correlated
placed with Cp by using the relation, and the last two terms of Eq. (75) must be replaced by

P = 1 + ---_--Cp (70) y;
to avoid underestimating the error contribution due to d.
The relative sensitivity coefficient, (M_/R)(/)R//_M**), is
given by Eq. (43) and is plotted in Figure 1 for _,=1.40.

Dynamic stability coefficients, Cmq and C/p


__gr_ _tnd moment coefficients, CF and Cm
Nondimensional dynamic stability coefficients are de-
Nondimensional force coefficients are defined by fined as derivatives of the moment coefficient with respect
F to a normalized angular rate. For example, the pitching mo-
C F = q**---_ (71) ment coefficient derivative, Cmq, is defined by

where F is an aerodynamic force, q** the free-stream dy- _m 0 (M/q**S/)


namic pressure, and S a reference area. The error in the Cmq /) (ql/V**) /) (q/IV**)
force coefficient is given by
_ /)M/Oq = ....Mq....

E__F_2
+ + ES 2 q**S/2/V** q**S/2/V
EC_ = CF I( F ; q**)
(Eq..12 ( -_- )11/2 (72) ** (76)
The derivative, Mq, is determined from the experiment.
where Er: is the error in the measured force, Es the error in Note that in these equations, M is the aerodynamic pitching
the reference area and Eq**is given by Eq. (46) as moment, whereas M.o is the free-stream Mach number, and
q is the angular pitch rate, whereas q**is the free-stream dy-

Eq**_q_[(M**
3R_ _,jEM_'2 . +_,( Ept)211/2pt
) .J [46] namic pressure" The estimated 2 err°r in Cmq is

where R = q**/ Pt, E M is the error in the free-stream Mach ECmq= Cmq Mq) +
EMq)

The relative Ep
number coefficient, (IVI._/R)(/)R/_gM_), is + (_Es _2 I__j/211
given byandEq. sehsitivity
is the'error
(49) in the measured
and is plotted in Figure 2total
for pressure.
_,= 1.40. "_" + _ (77)
Frequently, S = rid2/4, where d is a reference diameter.
As was noted in the previous section, if I = d and S = _12/4,
Then, l and S are correlated and the last two terms in Eq. (77)
should be replaced with (4Ed/d) 2. By definition,
Es_ = 25 (73)
S d
p vL p v
Nondimensional moment coefficients are defined by _ = 2V** = 2 (78)
C m = M (74)
q**Sl For a thermally perfect gas,

where M is an aerodynamic moment and l is a reference p._


length. The error in C m is given by P_ - R T (79)
g ,_

10
I

,i

and

V o,= M._ _,f_gT._ (80) Ec to= Clp i/ l


Lp) + _,_) 1
(41) to relate p_ to Pt and Eq. (43) of Ref. 12 to relate T** + + m (86)
Substituting these equations into Eq. (78) and using Eq. (Es)2 (2_/)2 ] 1/2
to Tt,
which is evaluated with Eqs. (61) and (82) as indicated

T--'_
= I+ M aboveforEq.(77).

gives M_ch number,

q,_ = 1 _ Pt M ( 1 + _M_1-2(--_(_-1) (81) In lfigh-speed wind tunnels, thefree-stream Mach num-


'2_Rg_t **\ ber is usually determined during tunnel calibration by one
of three methods, each involving a measured pressure
The relative error in q,JV** is given by ratio, R. First, at subsonic and transonic Mach numbers, a

Eq**/v** 2 2 0 R EM** p**.The measured tunnel total pressure, Pt, is used in the ra-
0 1,7
-** '.. = [(Eptl
Pt J + _,2Tt.
/ETt /) + ( _..__._
0M**I_I.0))21.] 1/2(82) staticR pipe
rio, maywith
= P**/Pt be used to measureflow
the isentropic free-stream
equation pressure,
to deter-
mine Mach number. From Eq. (41),
where R = q.. / V**.Differentiating Eq. (81) with respect to

R = -- = 1+ M 't-I [41]
M** OR 1 - M 2.* Pt
R 0M**= v_l _ (83) Solving for M**gives
1 + L_-_MZ 1

Comparing Eq. (83) with Eq. (61) shows that M - P_ -I_!_l - 1 (87)
kk Pt,)
M OR M. 0R a
R 0M_- RA 0M (84) Second, at supersonic and hypersonic Mach numbers, the
static pressure can be very low and the Mach number is
where R A= A,,./A*. Then, Eq. (61) and the results plotted usually determined with a pitot tube which measures the to-
in Figure 5 can be used to evaluate (M**/R)(0R/0M,,.) for
R = q**/V**.Equations (82) and (77) are used to estimate tal pressure, Pt 2 , behind the tube's normal shock. This pres-
the error in sure is divided by the tunnel total pressure, Pt, to give a ra-
Cmq" tio, R, which is related to Mach number by the normal

The rolling moment coefficient derivative, CIp, is de- shock relation, Eq. (53),
fined by,

0C I _ 0(L/q..S/) R = -- = ( )M_+2
Clp = _ (pl/V._) (9(pl/V)
1

_ q_S/2/V,
0L/0P = q. S/2/V.
Lp * (85) × I 2TM _- +1(_/- 1) l'/----i [53]

where L is the aerodynamic rollins moment and p is the roll This equation is solved numerically for M**,by using _mit-
rate. The derivative Lp is determin ;d from the experiment, erative root solver. Finally, a pitot-static probe may be used
The error in
Clp is given by to determine Mach number. In this case, the ratio and Mach
number dependency is given by the Rayleigh formula, Eq.
(56),

11
p

Rp-p,[ 2 l
rill2.(s121''2 (94)

r2yM_-- (y- 1)1_ -t where (M_/R)(aR/_M_)is given by Eq. (43).


x /L -y+l j [56] (2) For R = pt2/Pt the relative error in Mach number
is
Again, this equation is solved numerically for Moo.

= + (95)
fromThe
anyerror equation
of these for the Mach
three methods number
is the same. Letcalculated EM.*
M.* (___0R _-I
0--'M-_) [(Ept2)2
L_,Pt2 ) (Ept)21l/2
k Pt ) .J

Px ,,:',ere (M.*/R)(0R/0M**) is given by Eq. (55).


R (M.*,),) = -.- (88)
Py (3) For R = p.*/pt 2 , the relative error in Mach number
is
The uncertainty in th_ measured ratio is given by,

ER = Rr/_]
L_,Px) 2 + (_]21112
Py)J (89) EM,. (M____0R _-I [¢Ep../2 IEP,212],12
= k.R I.kP.*; +k %) J (96;}
Although an explicit relation for M**is not possible for two
of the ratios, the uncertainty in M** can be determined im- where (M.*/R)(0R/0M**) is given by Eq. (58).
• plicitly, when aR/0M.*# 0, by solving the error equatiolt,

OR Concludimz Remarks

E R "-a--_EM. (90)
Estimation of the uncertainty in results from high-
to give speed wind tunnel tests is hindered by the need to evaluate
the sensitivity coefficients, 3R/0M,., for various aerody-
0M**_ namic ratio,s, R. The ratios can be complex functions of M_
EM.* = R_-ff---ER (91) and calculating the partial derivative is a demanding task.
To simplify the error estimation procedure, sensitivity co-
Substituting Eq. (89) in Eq. (91) gives efficients were evaluated for five fundamental aerodynam-

EM.. R aM.. Ep,, condition (static or dynamic) to a reference (total or sonic)


M. - M, OR Lk_x
rl ]2 +k(Epy)2]
Py ) J 1/2 condition.
ic ratios. In Methodsof
theseapplying
general, ratios these asensitivity
relate free-stream coeffi-
test
cients in error estimation were demonstrated for several
For nonzero values of M**and OR/aM.*, nondimensional ratios and coefficients which are used in

f--"_-1 high-speed testing. In addition, the sensitivity coefficients

MR
•. OMoooR
= [M___
R 0--'_)0R
] (92) can be applied
free-stream directly toThe
conditions. estimating
presence the
of error in calculated
dependent or cor-
Then, related variables in a calculated test result is frequently a
challenge. Examples were given which compared two ana-

-- * = IMooaR ._-1
EM_ L_
r(Ep,/2_ + (Epy)2]l/2 (93) dependent.
iytical methods
Also,fora numerical
estimating method
the errorwhich
when approximates
variables are
Moo _--M_) ) PY ) J the partial derivatives was discussed.

Substituting the measured pressures for Px and py gives the


following estimates of error in calibration Mach number
for the three methods:

(l) For R =poo/pt the relative error in Mach number is

12
i

References

1. Abernathy, R. B., Thompson, J. W., Jr., et al,


"Handbeok - Uncertainty in Gas Turbine Measurements, '_
AEDC TR-73-5, Arnold Engineering Development Cen-
ter, Feb., 1973. Also published as: "Measurement Uncer-
tainty Handbook," Instrument Society of America, 1980.
2. "Measurement Uncertainty for Fluid Flow in
Closed Conduits," ANSI/ASME MFC-2M-1983, Ameri-
can Society of Mechanical Engineers, Aug., 1984

3. "Fluid Flow Measurement Uncertainty," ISO


TC30 SC9, Draft revision to ISO/DIS 5168, International
Standardization Organization, Dec, 1985.

4. "Measurement Uncertainty," ANSI/ASME PTC


19.1-1985, American Society of Mechanical Engineers,
April, 1986.

5. Coleman, H. W. and Steele, W. G., Jr., Experi-


mentation and Uncertainty Analysis for Engineers, John
Wiley & Soils, 1989.

6. Dieck, R. H., Measurement Uncertainty: Meth-


ods and Applications, Instrument Society of America,
1992

7. Clark, E. L., "Error Propagation Equations and


Tables for Estimating the Uncertainty in High-Speed Wind
Tunnel Test Results," SAND93-0208, Sandia National
Laboratories, Albuquerque, NM, August, 1993.

8. Coleman and Steele, op. cit., p. 80.

9. Moffat, R. J., "Using Uncertainty Analysis in


the Planning of an Experiment," Transactions of the AS-
ME, Journal of Fluids Engineering, Vol. 107, June 1985,
pp. 173-178.

10. Jones, P.A. and Friedman, M.A., "Propagating


Bias and Precision Errors Using the Perturbation Method,"
ISA Transactions, Vol. 29, No. 4, 1990, pp. 71-77.

l 1. Coleman and Steele, op. cit., p. 43 and Example


3.3 (p. 48)

12. Ames Research Staff, "Equations, Tables and


Charts for Compressible Flow," NACA Report 1135,
1953.

DISCLAIMER

This reportwas preparedas an accountof worksponsoredby an agencyof the UnitedStates


Government.Neitherthe UnitedStates Governmentnorany agencythereof,norany of their
employees,makesany warranty,expressor implied,or assumesany legal liabilityor responsi-
bility forthe accuracy,completeness,or usefulnessof any information,apparatus,product,or
processdisclosed,or representsthat its use wouldnot infringeprivatelyownedrights.Refer-
ence hereinto any specificcommercialproduct,process,or serviceby tradename,trademark,
manufacturer,or otherwisedoes not necessarilyconstituteor implyits endorsement,recom-
mendation,or favoringby the United States Governmentor any agencythereof.The views
and opinionsof authorsexpressedherein do not necessarilystate or reflect those of the
UnitedStates Governmentorany agencythereof.

13
i,
, m

Intentionally left blank

14
i,

m o

1,,2 .....
," _ • I ' ' ' ' " ' ..... i ' ' '

i
i aR/aM,
...... !
c_
_
o
0.4
"
iii
.

_ o.o
.s
s t•
s ;
t • °
sS •

-- 04¢ "_..................................................................
, " _ .............................................

a# #

-0,8 • ' ' i I I I I ,,I I ,,I I I _ I I, m ' •


0 2 4 6 8 10

Moo

a. R and OR / _M_
!

O , = , = ! , • 1 , , i i i i i

-2

-6

0 2 4 6 8 10

M=
b. 0 M = (M /R) (c)R#)M)
I

Figure 1. Variation of R, ;)R/0M_,, ,and 0 M,.,with Moofor R = Poo/Pt and y = 1.40

15
0.6
i ! i R
eo "_ - : ;

0.4

_E8 0.2
ct3

o_ o.o .
rY , ......:........
s ° :

-0.2 ................... _s '_,.'--'A@s S _ .........


.............................. _.............................................

-0.4 . . . , . . . t .... i . . . r . . .
0 2 4 6 8 10

M=

RandOR!OM_

-4

i : •
iiiiiii
iiiiii
: i

-6 ' • ' I , I , I . ' , I , J i I L. . •


0 2 4 6 8 10

M=
b. 0"M _ = (M JR) (0R/0M)
i

Figure 2. Variation of R, OR/0M_, ;u_d 0 M with Moofor R ---q_/P, _md y = 1.40

16
R_ aR/aMoo
(MJR)(aR/aMJ ,
0
,
0 ¢:) 0 0 ...,

e-

o old,..,
i,.........i)"
P.

5.
_- • : "
_" _ ... _ ............ , .................................................. : ..............
= \i ? i

8 " .I_ *- I_ .1_ ..... :........... .,-................ ;................ :...............

_ _ _ ...........................................................................

i
II

_ 1_ I_ ...................................................................

a2

II

•1_ C) 0
. i

1.2
i i i R ......

0.4

o.o ............................ .--, I


t

i tI s sIS
t •

-0.4 -_ .......... _'...................................................... _.............................................


t •

, I /
_ SG

-0.8 . . . I i i , I , , , I i , i I . • .
0 2 4 6 8 10
M_

a. R and _R / _M_

0.0

-0.4

"_ -0.8

8 -1.2

-1.6

-2.0
0 2 4 6 8 10

b. 0 M = (MJR) (OR#-)M)
t

Figure 4 V_uialionof R, _R/c)Moo,and 0 M with Mootbr R = p_,/pt2 _mdy = .'.40


i
b .i

6OO
"" R .....

400 ..................... _....................... _..................... )...................... _ ..............

• : ; : ss S

200

.....e.± ,

_200 i . . . , . . . I . . . I . . . I . . .
n 2 4 6 8 10

a. R and OR/OMoo

, _ i, i i e i , • _ , e v , ijllll iii , e e

4 .....................
8
ct_

c_

_ 2 .....................
8

_ 0

-2 , , I I I ' ' I ' , t I , , ,.. I , ' '


0 2 4 6 8 10

Moo
b. 0'M ' = (M JR) (o-)R/c--)Moo)
--
t

Figure 5. Variation of R, OR/OMo.,


and 0 M=With Mo, for R = Ao,/A* and T= 1.40

19
i
e

ADDENDUbl

Sensitiviw Coefficiems for Ratio of S0ecific Heats.

A colleague recently suggested that I should exam- The relative error equation for free-stream dynamic
ine the sensitivity coefficients for the ratio of specific heats, pressure is the same as that given above for static pressure,
"t.In the main body of the paper, it is assumed that the error except that R = qoo/P,,

due
whereto 2'is
real negligible,
gas effects i.e.,
are E_,
= 0 At and
significant hypersonic
errors inspeeds,
), are
probable, the sensitivity coefficients for _,are large. There-
fore, any errors in 2,will be magnified.
Eq.
__ =
qo. I/ M._ DR E M
R /)M**M_') + kR _,
+
/ _) _1
The sensitivity coefficients for q.. are given in Figure A-2.

The error equation for free-stream static pressure, The significant influence of errors in T at hypersonic
p.., is, from Eq. (33), Mach numbers is evident from the plotted coefficients. For

0.07%) wi:! result in relative errors of approximately 0.8%


Ep. = [/3p.o E M..)/2
[._0-M--_ + (_** E'I12 +_--P'tt
//)p.E Pt)j
)211/2 example,
in p.o and atM**=8, an error in 'yof 0.001(relative error of
0.9% in q**.

Expanding to the relative error form gives

p"_ = R aMooM_) + 27 _/J + "_'t} J


where R = p,_/Pt. The absolute and relative sensitivity co-
DR , _3R
efficients, 0y = _-_ and 0y = R_'T' respectively, were
evaluated using numerical differentiation (Eq. (10) with
Ay/y = 0.004 ) and are presented in Figure A-I.

A-I
t
, J

1.00

0.75 ...........................................................................
! i i i

cD 0.50

n,," 0.25 ............... '_--- -- '............................................

0.00 ..........
• ,, .

-..o

-0.25
0 2 4 6 8 10
Moo

& RandOR/_/

i , , , • ,

"- ' ' ' '' ..... ' ' I

12

,
.&-,
_ 4

0 -____.._..._

--4
0 2 4 6 8 10
M

b. O _f (y/R) (onR/ony)

Figure A-I. Variation of R, aR/O),,and O's,with M**for R = P*_/Ptand 1,= 1.40

A-2
m

# ql
• q

6 ! , ! ! ! ! i i • _11 ! , , iii I ' , i iiiiiiif

4 ......

0 ...... i _
0 2 4 6 8 10

b. 0 _ = ('C/R)(aR/a¥)

FigureA-2. Variationof R, aR/_, and 0'1.wilhM_ for R = Cl_/Ptand'y= 1.40

A-3
m m
I I
I

Вам также может понравиться