Professional Documents
Culture Documents
Key words : wind tunnel test, uncertainty, technological The Aerodynamic Tunnel no. 2 (TA-2) of the Centro
inovation. Técnico Aeroespacial (CTA). The TA-2 has been in service
for over thirty years and has been used for many aerospace
projects, mostly Embraer aircraft. The subsonic wind tunnel
1. INTRODUCTION TA-2, which is the largest in Latin America, has a test
As the commercial aircraft industry deals with complex section measuring 2.1 by 3.0 meters (7 by 10 ft). (Figure 1).
products which encorporate the integration of several There are two laboratories, the Low Pressure Laboratory and
subsystems, traditionally there are long periods from the the Low Force Laboratory, connected to TA-2, which
design of new products up to their efective final market provide support for its activities, carrying out calibrations
delivering. In recent years, the industry, in general, has on sensors used in aerodynamic testing.
sought to increase design efficiency, coupled with time and
requirements are very strict and the test-section turbulence
level is to be as low as 0.05% of the mean flow kinetic
energy. To insure that the desired turbulence level and flow
uniformity at the test section is achieved the tunnel will have
a 10:1 contraction, a honey comb and three screens with
provision for an extra one. Furthermore, the 35-meter long
facility has its inlet section, the contraction and the test
section inside a 25x10 room. The diffuser, the fan section
and the exit section are mounted outside the laboratory
room. Figure 3 shows the tunnel layout.
The ASA-L staff has been preparing the Low Pressure Figure 4: Bi-dimensional study at the USP/São Carlos wind
Laboratory and the Low Force Laboratory to take part in the tunnel.
Brazilian Calibration Network. Being accredited laboratories
will bring them formal recognization of competence.
Along with this series of tasks to achieve levels of services 3. METHODOLOGY
similar to international ground test facilities, it is the
adequate estimation and assessment of the uncertainty in 3.1. Improvement in the productivity
measurement, which allows the comparison of test results
and contributes to the metrological reliability of the The commonly quoted metrics for wind tunnel productivity
Brazilian Aerodynamic Laboratory. are:
2.2. The ITA Wind Tunnel − Polars per hour of occupancy time [1]. Polar is used in
ground testing terninology to test a model in a determined
A non-return wind tunnel is currently under design at the configuration and test condition. A polar is acquired by
Instituto Tecnológico de Aeronáutica (ITA). The support is taking readings for several angles of attack, in fan-on
being provided by this project. The test section flow quality condition;
− Polars per fan-on hour;
− Rate fan-on hour to hour of occupancy; 3.3. Uncertainty in the loads
− Percentage of the facility down time (due to facility
problems); The loads Fi (i = 1,...,6) act on the model during the wind
− Time to model access after accomplishing a polar tunnel test; F1 , F2 and F3 denote forces ( drag, side and lift,
acquisition (time from fan stop to hands on the model ); respectively) and F4 , F5 and F6 denote moments (rolling,
− Time to change the model configuration (get-the-model- pitching and yawing, respectively). The external balance
change-done –fast); measures the loads. The calibration of the balance
− Time of wind-on availability to obtaining data. contributes greatly to the uncertainty of measurements in
wind tunnels.
In order to optimize the parameters listed above, the In order to obtain a suitable model to provide aerodynamic
upgrades of the following systems are being conducted: loads from the balance load cell measurements, a balance
calibration is performed prior to the tests. The calibration is
accomplished by applying loads to the balance through a
− Drive system (motor and fan) ;
system of cables and pulleys named calibration cross
− Velocity control;
(Figure 5).
− Data acquisiton hardware; Seventy-three loading combinations are used. The loads F1 ,
− Data acquisiton software; and F2 , F3 , F4 , F5 and F6 applied to the system are measured by
− Model positioning. six load cells, which supply the readings R1 , R2 , R3 , R4 , R5
and R6 .
3.2. Improvement in the reliability
The TA-2 flow has being studied over the years and
improvements have being planned such as the fan changing,
the treating of the internal tunnel surface to minimize the
wall boundary layer separation and to prevent points of
leakage. Other steps may be adopted such as small
corrections in the corner vanes and the introduction of
screens.
The uncertainty in measurement must be evaluated and
documented [3]. The major error sources, in relation to the
final results, must be identified and the errors minimized. In b)
this way, the Aerodynamic Laboratory staff has conducted
several studies in order to identify the contribution of Figure 5: Loading system for balance calibration: the
uncertainty components [4] and to develop methodology of calibration cross. a) picture. b) schematic. The numbers
the assessment of the aerodynamic external balance 1,...,14 represent points of application of weights.
calibration uncertainty [5,6]. In this inovation project, this
work is to be continued, with the implementation of the A calibration curve is fitted by the least squares method to
methodology of evaluation of the uncertainty in the each set of 73 points (Fi ; R1 ,R2 ,R3 ,R4 ,R5 ,R6 )k, i = 1,...,6 [6]:
reduction system.
Fi (R) = a 1 R1 + a 2 R2 + a 3 R3 + a 4 R 4 + a 5 R5 + a 6 R6 + model which relates the forces and moments to the load
cells readings provided by six-component external balance
a 7 R + a8 R1 R2 + a 9 R1 R3 + a 10 R1 R 4 + a11R1 R5 + a12 R1 R6 + (Equation 1). The variances, the covariances and the non-
1
2
a13 R22 + a14 R2 R3 + a15 R 2 R4 + a16 R2 R5 + a 17 R2 R6 + a18 R32 linearity of the model were considered.
There are several error sources presented in the balance
+ a19 R3 R4 + a 20 R3 R5 + a 21 R3 R6 + a 22 R42 + a 23 R4 R5 + calibration. Some of them are recognizable and others may
remain unknown. The identified error sources were
a 24 R 4 R6 + a 25 R5 + a 26 R5 R6 + a 27 R6
2 2
quantified and considered in the uncertainty assessment [6].
(1) The results of the methodology applied to a specific set of
The fitting model is a linear combination of 27 (twenty calibration data points are shown in Table 1. The quantities
seven) functions of R, called the basis functions, which are: Fiapplied , i = 1,…,6, are the loads applied to the calibration
2 2
R1 , R2 , R3 , R4 , R5 , R6 , R1 , R1 R2 , R1 R3 , R1 R4 , R1 R5 , R1 R6 , R2 , cross. The quantities Fifitted correspond to the value of the
2 2
R2 R3 , R2 R4 , R2 R5 , R2 R6 , R3 , R3 R4 , R3 R5 , R3 R6 , R4 , R4 R5 , loads estimated through the curve fitting. The uncertainty in
2 2
R4 R6 , R5 , R5 R6 , R6 and has twenty seven adjustable the loads, u(Fi ), was estimated using Equation 3.
parameters a j (j =1,...,27) for F1 , twenty seven adjustable The coverage factor used to evaluate the exapanded
parameters b j for F2 , and so on, until reaching the fj uncertainty, U(Fi ), is derived from the effective degrees of
parameters for F6 . The model’s dependence on its freedom υeff , estimated from the Welch-Satterthwaite
parameters a j ,b j ,...,fj ,(j =1,...,27) is linear. The functions of formula [3].
R, are nonlinear.
The equation for the combined uncertainty of the estimation Table 1: Estimation of uncertainty in the Fi component.
of the measurand y, for N correlated input quantities and for (Unit: newton for force and newton×meter for moment).
a nonlinear model is the positive square root of [3]:
2 Fi u(F1 ) Fi Fi
∂y 2 ∂ y ∂y
( )
N N −1 N
u ( y) = ∑
2
u (xi ) + 2 ∑ ∑ u xi , x j + applied fitted
∂ xi i =1 j = i +1 ∂xi ∂x j
c
i =1 F1 0.39 0 0.51
(2)
F2 0.38 0 -0.20
N N 1 ∂2 y ∂y ∂ 3 y 2
2
+ ∑∑
∂ ∂
+
∂ ∂ ∂ 2 u ( xi )u x j
2
( ) F3 1.17 -2400 -2400.12
i =1 j =1 2 x x x x x
i j i i j
F4 0.29 0 -0.51
F5 0.37 -300 -299.77
Applying the law of propagation of uncertainty (2) to the F6 0.35 0 0.02
equation which represents the fitted model (1) leads to the
combined standard uncertainty of the values Fi , estimated
by the curve fitting, for any readings Ri : For the lifting force, F3 , the reported result shoud be
declared as:
u c2 (F i ) = (R1 ) u 2 (a1 ) + (R 2 ) u 2 (a 2 ) + K
2 2
“F3 = (2400.1 + 2.4) N. The reported expanded uncertainty
is stated as the standard uncertainty multiplied by the
+ ( R6 ) u 2 (a 6 ) + K + (R6 ) u 2 (a 27 ) +
2 4
coverage factor k =2.04, which for a normal distribution
+ 2[ R1 R2u (a1 , a2 ) + R1 R3 u (a1 , a3 ) + L corresponds to a coverage probability of approximately
95%.”
+ R5 R6 R62u (a 26 , a 27 )] +
+ ( a1 + 2 a 7 R1 + a 8 R2 + a 9 R3 + a10 R4 + 5. CONCLUSIONS