JOURNAL OF SPACECRAFT AND ROCKETS

Vol. 46, No. 3, May–June 2009

Optimal Aft End Distorted Fin Model Using

Response Surface Method

Se-Yoon Oh∗
Agency for Defense Development, Daejeon 305-600, Republic of Korea
and
Seung O. Park†
Korea Advanced Institute of Science and Technology, Daejeon 305-701, Republic of Korea
DOI: 10.2514/1.40496
In wind-tunnel testing of bomb or missile-type models, aft end geometry distortion may become necessary when a
model is tested by using a sting support system. When the diameter of the model base is increased, the exposed tail fin
area is reduced to result in aerodynamic data alteration. In the present work, various tests are conducted on an MK-
82 bomb model with base diameter and exposed tail fin area variations. The tests use techniques involving design of
experiments and response surface modeling. The effects of the model base geometry modifications on the
Downloaded by BEIHANG UNIVERSITY on July 1, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.40496

aerodynamic characteristics are investigated. Experimental design optimization is then performed on the basis of
three design factors: the base diameter, the extended tail fin area, and the freestream velocity. The experimental
results show that the altered aerodynamic characteristics due to aft end model distortion can be minimized by
changing the tail fin geometry on the basis of the regression model.

Nomenclature aerodynamic interference than vertical support mounts, such as belly

A = extended tail fin area ratio, S0t =St supports or blade struts. When a test model does not have a base area
Cm
= pitching moment coefficient slope large enough for sting support due to restrictions of the wind tunnel, it
C N
= normal force coefficient slope may be necessary to modify the geometry of the model aft end.
ctail = tail volume coefficient Usually the model base diameter is increased so that the effective
D = reference body diameter or maximum diameter surface area of the tail fin is reduced; however, this type of reduction
Db = base diameter of the model means that the aerodynamic characteristics of the test model differ
Kb–t = body–tail interaction factor from those of an undistorted model.
Kw–b = wing–body interaction factor The model aft end distortion corrections of aerodynamic data are
Lt = tail moment arm, Xcg
Xcpt quite different from the usual support interference corrections [1].
ReD = Reynolds number based on the reference body Nonsimilar geometric change in the aft end of the model makes the
diameter correction more difficult; and additional tests on an undistorted
S = missile reference area, D2 =4 model are required for reference data. Although various works have
St = exposed tail fin area focused on the sting interference correction of flow distortion, aero-
Sw = exposed wing area dynamic correction techniques for geometric model distortion are
S0t = extended St rare because most wind-tunnel tests assume geometric similarity
Ur = 95% confidence uncertainty between the model and the prototype. An example that convincingly
Xcg = center of gravity location from nose tip demonstrates the need to study geometrically distorted model testing
Xcp  = center of pressure location from nose tip is the store model test of a captive trajectory system. The external

= angle of attack, deg shape modification may be compulsory due to the conflicting re-

Cm
= pitching moment coefficient slope change quirements of a scaled-down model (typically, 0.05-scale models are

CN
= normal force coefficient slope change used) and the sting support [2,3]. Aerodynamic data that are modified
= angle of yaw, deg as a result of model aft end distortion are likely to cause a substantial
change in trajectory [2]; hence, if available, undistorted aerodynamic
data should be used on account of the geometric distortion of the
model.
I. Introduction To establish a reliable model modification and compensation
HIS study aims to find a way to minimize distortion effects on process for the distortion effects, we first need to determine the effect
T aerodynamic data due to the inevitable model distortion caused
by sting support in wind-tunnel testing. Most missiles and stores are
of the tail fin design parameter on aerodynamic characteristics. In the
present work, we considered three factors as design variables for
bodies of revolution, and their wind-tunnel test models are usually the purpose of developing a compensation technique for the aft end
handled with sting mounts because the sting mount produces less distortion problem of a model of the MK-82 bomb. The three factors
are the test velocity (V), the aft end distorted model base diameter
Received 19 August 2008; revision received 18 November 2008; accepted (Db ), and the extended area ratio of the tail fins (A) based on the fin
for publication 23 December 2008. Copyright © 2008 by the American planform area. We chose the design of experiment-based response
Institute of Aeronautics and Astronautics, Inc. All rights reserved. Copies of surface methodology (DOE/RSM) [4,5] to investigate the effects of
this paper may be made for personal or internal use, on condition that the the tail fin design. The DOE/RSM technique has been widely used in
copier pay the $10.00 per-copy fee to the Copyright Clearance Center, Inc., recent wind-tunnel experiments [6–14], and a number of studies that
222 Rosewood Drive, Danvers, MA 01923; include the code 0022-4650/09 use the DOE/RSM technique for wind-tunnel tests have been
$10.00 in correspondence with the CCC.
∗
Senior Researcher, Wind Tunnel Test Division, Yuseong Post Office Box
published. DeLoach et al. [6–10,12] have made significant contrib-
35-12; also Ph.D. Student, Korea Advanced Institute of Science and utions in this regard.
Technology, Daejeon 305-701, Republic of Korea. The first part of this paper investigates the nature of the model aft
† end distortion problem and introduces the notion of the equivalent
Professor, Department of Aerospace Engineering, 373-1 Kusong-dong,
Yuseong-gu. Senior Member AIAA. tail volume to deal with this undesirable contour alteration. The
592
 OH AND PARK 593

speed wind tunnel to obtain aerodynamic data for various tail fin

configurations. The tunnel is a solid-wall single-return closed-circuit
tunnel with a 9:1 contraction. The test section has an area of 3 
2:25 m and a length of 8.25 m. The average axial turbulence level of
the tunnel is about 0.08%, and the nonuniformity of the tunnel flow
angle is less than 0.1 deg. Aerodynamic forces and moments were
measured with an internal six-component strain gauge balance at a
sampling frequency of 10 Hz, and the data were averaged over 50
samples. An MC-10-1.50-A internal six-component strain gauge
balance (manufactured by MicroCraft Technology, San Diego,
Fig. 1 Distorted MK-82 model and undistorted model with image
supports.
California) was used. The load ranges of the balance are as follows:
normal force up to 1000 lbf and pitching moment up to 2000 in:  lbf.
The overall accuracy of the balance is within 0:1% of the full scale.
second part focuses on the DOE/RSM technique as a means of Figure 1 illustrates a scaled-down MK-82 distorted fin model
compensating for the distortion effects. Finally, there is a presen- mounted on a crescent arc system with a rear sting support. A forced
tation of the experimental results for the various distorted tail fin transition technique [15] was adopted to ensure a turbulent flow.
designs based on the DOE/RSM technique. Transition dots of 0.28 mm (0.011 in.) in height were attached to the
nose at 5% of the body length position from the nose tip and on the
tail fins at 15.7% of the chord from the leading edge. As shown in
II. Experimental Details Fig. 2, the present MK-82 model has characteristics that are very
Downloaded by BEIHANG UNIVERSITY on July 1, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.40496

For many scaled-down experiments on unguided, nonpropelled common in missile and bomb designs, namely, a tapered afterbody
bomb-type configurations, the diameter of the sting in the test model and cruciform tail fins [16]. The tail fin has a tapered planform, as can
may be greater than the diameter required by the geometric similarity be seen in Fig. 2.
rule, especially if the sting has to withstand experimental conditions Under geometric model discrepancies [2], the variation of the
such as tunnel starting loads and maximum steady-state loads [15]. In longitudinal static stability derivatives, such as the CN
and Cm

these circumstances, the aft end shape of the model should be made values of a slender-finned bomb at subsonic Mach numbers, is much
large enough to accommodate the tail sting support, and this increase larger than the corresponding variation at supersonic Mach numbers.
in size causes the aft end model to be distorted. The present work We therefore confine our attention to low subsonic flow cases. A
concerns the way the aerodynamic data is altered due to distorted simple equation for a normal force coefficient derivative of a missile-
aft end models of the 50% scale MK-82 bomb with fixed cruciform type body is given as follows [16], where the suffixes n, b, w, and t
(x-type) tail fins at low subsonic speeds. As the main purpose of this denote the contributions, from the nose, body, wing, and tail,
work is to investigate the effects of aft end distortion, we simply respectively:
selected nine different distortion cases (that is, three base diameters
by three extended fin areas). Experiments were carried out in a low- CN
 CN
n  CN
w Kw–b Sw =S  CN
t Kb–t St =S (1)

Fig. 2 Model configuration of the 50%-scale MK-82 model with extended tail fins and model bases.
 594 OH AND PARK

Note here that the wing and tail area are the exposed areas; thus, in
contrast to the planform area used in airplane aerodynamics, they do
not include the area immersed in the body [16,17].
The pitching moment equation about the c.g. is expressed as
follows [16]:


Xcg
Xcpn  Xcg
Xcpw  Sw
Cm
 CN
n  CN
w Kw–b
D D S


Xcg
Xcpt  St
 CN
t Kb–t (2)
D S

where various positions of center of pressure (Xcp ) are dominant

factors. We assumed that the location of the center of pressure of the
tail was at 25% of the mean aerodynamic chord (0.25 MAC) of the
tail fin. Fig. 3 Pitching moment curves for various MK-82 model aft end
The effectiveness of the tail in generating moment about the c.g. is conditions.
proportional to the product of the normal force of the tail and the tail
moment arm as given in Eq. (2). For this reason, the tail effectiveness
is usually expressed by the tail volume coefficient as follows:
Downloaded by BEIHANG UNIVERSITY on July 1, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.40496

tail volume Xcg
Xcpt  St Lt St

ctail    (3)
reference volume D S DS
Given the significance of the pitching moment in the dynamics
of a missile or bomb-type body, the tail fin area should be changed
S0t  so that the tail volume coefficient remains unchanged. For
simplification of the aft end distortion, it may be better to keep Lt
(see Fig. 2) constant. For constant Lt , we need to change the tail fin
geometry so that the location of 0.25 MAC of the tail fin does not
change. Accordingly, both the aspect ratio (ARt ) and the sweepback
angle (t ) should be changed as shown in Fig. 2. With Lt constant,
Eq. (3) tells us that we should change the tail fin geometry so that the Fig. 4 FCCD and CCD for the DOE/RSM experiments (coded values).
exposed fin area remains the same as that of the (original) undistorted
fin, thereby ensuring that the tail volume coefficient remains
unchanged. However, this process is known to cause a significantly curve of the undistorted model [4–14]. We therefore resorted to the
distorted Cm
, partly due to the modifications of CN
t and Kb–t after DOE/RSM approach.
the geometric distortion. Thus, an aerodynamicist must decide which The central composite design (CCD) [4–7,9,10] is the most
fin area would be the best for minimizing the distortion in Cm
. popular class of designs used for fitting a second-order regression
Figure 3 presents pitching moment curves for various extended tail model. In the present study, a DOE with a face-centered composite
fin cases from our preliminary experiment. An independent wind- design (FCCD) [4,5,13,14] is applied to the RSM to determine how
tunnel test was conducted to acquire data from an undistorted model the tail fin design parameter affects the aerodynamic characteristics
(dashed line in Fig. 3) for reference. To evaluate and remove the strut (Fig. 4). The experimental design variables in this work are the base
support interference effects, we installed a dummy, or image, strut diameter, the exposed tail fin area, and the test velocity. Three
support that is identical to the original strut support shown in Fig. 1. distorted model base diameters and three tail fin extended areas were
The use of image support for the correction of the interference selected to account for the effect of the aft end distortion. Three test
effect is well known [1,15]. For the distorted model tests, the rear velocities were chosen to examine the effect of the Reynolds number,
sting interference effects were accounted for in terms of the flow even though the effect is not expected to be as significant as the
angularity correction runs (upright and inverted runs) and the normal force and the pitching moment is essentially independent of
additional
Cm correction from the pitching moment coefficient viscosity at low angles of attack. The independent circumscribed
difference in the upright and inverted runs [15]. The uncertainty CCD experiments (Fig. 4) on the
and combinations (with an

estimates of various measurements of the present work were carried and range from
5 to 5 deg, corresponding to 1 in the coded
out in accordance with AIAA standard S-017A-1999 [18]. The value) were conducted for various aft end distorted models. The three
uncertainties in the pitching moment coefficient for the undistorted design factors with three levels are summarized in Table 1. In the
model were found to lie between 0.0584 at
 0 deg and 0.0586 at FCCD test matrix, the factor of interest is denoted in the coded level

 5 deg. Figure 3 clearly illustrates the need to change the (linear transformed level) given in Table 1.
exposed tail fin area so that the effect of the aft end body distortion The least-squares regression technique is usually applied to fit the
can be properly explained. We see that the case of a 2% area RSM. This regression [4,5,7] is quite different from an ordinary
extension yields a pitching moment curve closest to that of the un- regression because the regression in the DOE/RSM is a predefined–
distorted model. However, in an experiment based on the traditional designed regression for a DOE approach. This DOE/RSM-based
one-factor-at-a-time approach, it is hard to determine the exact regression provides us with mathematical response models of longi-
extended area required to correctly produce the pitching moment tudinal static stability derivatives as a function of all the design

Table 1 Factor levels for the FCCD experiments

Natural level (i ) Coded level (xi )
Name Factor Units Low Center High Low Center High Remark
Base diameter Db mm 65 85 105
1 0 1 0:48 Db =D 0:77
Extended fin area A % 2 6 10
1 0 1
t  2:9
,
ARt  0:12
Test velocity V m=s 40 60 80
1 0 1 3:6  105 ReD 6:9  105
 OH AND PARK 595

Table 2 FCCD matrix for experiments and measured results

STD No. Run No. Db , mm A, % V, m=s
Cm
, 1/deg
CN
, 1/deg
15 1 85 2 60 0.0091 0.0007
17 2 85 6 40
0:0355 0.0112
13 3 65 6 60
0:0569 0.0157
16 4 85 10 60
0:0957 0.0240
20 5 85 6 60
0:0443 0.0133
19 6 85 6 60
0:0453 0.0142
14 7 105 6 60
0:0161 0.0061
18 8 85 6 80
0:0513 0.0145
6 9 85 6 60
0:0454 0.0134
4 10 65 10 80
0:1188 0.0307
3 11 105 2 80 0.0386
0:0077
5 12 85 6 60
0:0457 0.0138
1 13 65 2 40
0:0006 0.0030
2 14 105 10 40
0:0641 0.0174
11 15 85 6 60
0:0465 0.0150
8 16 65 10 40
0:1030 0.0272
7 17 105 2 40 0.0411
0:0060
12 18 85 6 60
0:0414 0.0122
1 19 105 10 80
0:0757 0.0196
Downloaded by BEIHANG UNIVERSITY on July 1, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.40496

9 20 65 2 80
0:0114 0.0063

factors in Table 1. The regression models used in the present study III. Results and Discussion
are of a second-order form as given in the following equation: A. Experimental Data and Analysis
The parameters Cm
and CN
of various aft end distorted models
X
k X
k X
k
1 X
k
were obtained by varying
and as specified in the CCD matrix of
y  0  i xi  ii x2i  ij xi xj  " Fig. 4. These slopes for the distorted and undistorted models were
i1 i1 i1 j2
obtained by using a least-squares fit for each data set (eight
i
maxi   mini =2 data points). The parameters
Cm
and
CN
with respect to the
for n observations where xi  undistorted model data were then calculated. The test results are
maxi 
mini =2
(4) summarized in Table 2. The uncertainties [18,19] in the aerodynamic
force and moment coefficients for the distorted model are
summarized in Table 3. The order of all the test runs was randomized
where each  is a model coefficient to be determined empirically; to average out the influence of unknown variables over time. The
each x is a regressor; y is a response variable, such as the pitching tests were repeated six times about the center point (Db  85 mm,
moment coefficient; and " is a normally and independently dis- A  6%, and V  60 m=s) of the FCCD matrix. Repeated runs of the
tributed random error or residual assumed to have a zero mean value same test achieved a repeatability value of Cm
 0:00180= deg.
and a constant variance. The variable xi in Eq. (4) transforms the The Minitab [20] and the Design-Expert [21] software packages
natural or actual variable i so that the value falls between
1 and 1 were used for data processing of the CCD matrix generation,
as in Table 1. regression analysis, analysis of variance (ANOVA), and response
optimization. Table 4 gives regression analysis results of
Cm
in
coded units. The coefficients represent the mean change in response,
Table 3 Estimated uncertainties for a distorted model
(Db
85 mm, A
6%, and V
60 m=s)
Cm
, for one unit of change of the corresponding factor while
holding other factors constant. The p value in the table for each
Response Test results with uncertainties Name regression coefficient tests the null hypothesis H0 [4,5]. In general, a
CA 0:1766  0:0509 Axial force low p values suggests that the corresponding regressor variables
CY 1:0356  0:1240 Side force (xi ) contribute significantly to the regression model. However,
CN 1:0508  0:1462 Normal force by comparing the t-for-H0 values of Db , A, and V at a 5% level of
Cl 0:0384  0:0394 Rolling moment significance, we see that Db and A are much more significant
Cm
4:5371  0:0639 Pitching moment variables than V. We also find that the quadratic term D2b is more
Cn
3:9315  0:0351 Yawing moment important than the other quadratic terms and interaction terms, even

6:1629 deg 0:0348 deg Angle of attack though its effect on the regression model is not critical. The re-
4:9968 deg 0:1119 deg Angle of yaw
gression results given in Table 4 indicate that the response surface is

Table 4 Estimated regression coefficients for
Cm
(R2
99:88%, R2 pred
99:33%,
R2 adj
99:77%)
Term Coefficient Standard errors t for H0 coefficient  0 P value Significance (P < 0:05)
Constant
0:044320 0.000690
64:226 0.000 Significant
Db 0.021450 0.000635 33.792 0.000 Significant
A
0:053410 0.000635
84:142 0.000 Significant
V
0:005650 0.000635
8:901 0.000 Significant
D2b 0.007150 0.001210 5.907 0.000 Significant
A2 0.000350 0.001210 0.289 0.778 Not significant
V2 0.000250 0.001210 0.207 0.841 Not significant
Db A
0:001213 0.000710
1:709 0.118 Not significant
Db V 0.001562 0.000710 2.202 0.052 Not significant
AV
0:001763 0.000710
2:483 0.032 Significant
 596 OH AND PARK

approximately linear in Db and A. The various R2 values (coefficient

of determination) imply that the present regression model fits the
data extremely well. An ANOVA table (testing for a lack of fit) for

Cm
, which is not presented here, also indicates that the regression
model adequately describes the measured
Cm
data. The regression
analysis results of
CN
are similar in character to those of
Cm
;
thus, we omitted the data here.
To guarantee the validity of the regression model, we must always
ensure that none of the regression assumptions [4,5] are violated. For
this purpose, we checked the normality assumption by constructing a
normal probability plot of the residuals; the results are shown in
Fig. 5. We see from Fig. 5 that the residuals are scattered around a
straight line, which implies that the normality assumption is satisfied
at a 5% level of significance. We also confirmed that the present
regression curves do not violate the constant variance assumption
and that the residuals are not correlated with each other.
Figure 6 shows a response surface plot and a contour plot of
Cm

as functions of Db and A. In this figure, the
Cm
data from the Fig. 6 Response surface and optimization solutions for
Cm
; velocity
measurement (20 data points) are marked with circles. The dashed- at 60 m=s.
line curve on the response surface of the figure is discussed in
Downloaded by BEIHANG UNIVERSITY on July 1, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.40496

Sec. III.B. To check the adequacy of the response model, we tested an

extra tail fin model (Db  85 mm and A  4%) as a confirmation
experiment. The measured value was compared with the prediction The design factors and object responses of the response optimi-
value from the regression model. Using our response model, we zation are listed in Table 5. For the purpose of presentation, we first
find that the 95% prediction intervals for this confirmation point set the test velocity to 60 m=s and the model base diameter to 84 mm
are
0:0220
Cm

0:0130 and 0:0045
CN
0:0096. as goal point values. We then search the optimum extended fin
The actual measurement yielded values of
Cm

0:0141 and area that yields the maximum value of the desirability function.

CN
 0:0070, which clearly demonstrates the accuracy of the Obviously, the goal points for
Cm
and
CN
are 0 whenever
present response model. distortion-free conditions are desired.
Figure 7 shows the optimization plot. The vertical dashed-dotted
lines on the graph and the numbers displayed at the top of each
B. Response Optimization column represent the current design factor settings (Db  84 mm,
This section concerns the question of how to select an extended fin A  2:62%, and V  60 m=s). Various curves show how the de-
area that minimizes the aerodynamic distortion when a specific sirability functions dcomp ,
Cm
, and
CN
vary with respect to the
(distorted) base diameter and velocity are given. The dashed line in variable of each column whereas the other variables are kept constant
Fig. 6, which is the curve of
Cm
 0 on the response surface, at their optimum values. We again see that the extended fin area
graphically illustrates the situation. When the base diameter is should be 2.62% when Db  84 mm and V  60 m=s. In this case,
84 mm, we find from the curve that the extended fin area should be the composite desirability, dcomp , becomes 0.95103 whenever
CN

2.62% to give a zero-value
Cm
. When there are multiple response is different from 0 with d2 0:905.
functions, a more systematic approach is required. In terms of the An experimental aerodynamicist would be more interested in
response surface method, that means finding the best factor or factors finding a suitable extended fin area when the base diameter has to
that yield an optimum response in which the desirability function be distorted by a certain amount. Figure 8 shows an overlay plot
[4,5] is used. The desirability function, di , is an objective function generated from the regression model for this purpose. The shaded
that ranges from 0 outside of the validity limits to 1 at the goal point. area represents the range of the extended area that yields values of
We use a composite desirability function [4], namely, dcomp 
Cm
 0 1 standard error in prediction (1SE  0:0020= deg
fd1  d2      dm g1=m , which refers to the weighted geometric in
Cm
corresponding to 0:3% of the pitching moment slope of
mean of the desirabilities of the individual responses, where m is the the undistorted model). For example, we easily find that an extended
number of responses in the measure. In the present case, we set the fin area of 2% is needed when the base diameter is 75 mm. When the
composite desirability function as follows: base diameter must be greater than that of the undistorted model, the
tail fin area is inevitably immersed. From an aerodynamic viewpoint,
dcomp  fd
Cm
  d
CN
g1=2 (5) we are interested in the amount of exposed fin area needed to properly
compensate for the immersed area so that, in spite of this geometric
distortion, Cm
does not undergo any change. Figure 9 shows the
relation between the extended area and the immersed area. When the
base diameter is varied from 65 to 105 mm, the immersed area based
on the fin planform is increased from 7 to 21% in the present case.
Figure 9 shows that the extended fin area varies from 1.54 to 4.85%
with an increase of the immersed area. We can easily see that the
necessary extended area is approximately 24% of the immersed fin

Table 5 Parameters for response optimization

Factor (x) Goal Lower limit (L) Upper limit (U)
Db Is equal to 84 65 105
A Is in range 2 10
V Is equal to 60 40 80
Response (y)

Cm
Is target T  0 - -
Fig. 5 Normal probability plot of residuals with a 95% confidence
CN
Is target T  0 - -
interval for
Cm
.
 OH AND PARK 597

area. This result is consistent with the observation made in the

computational study of [22] for the MK-84 (Joint Direct Attack
Munition). This paper showed that, with a large sting, the area-
matched fin configuration (that is, when the exposed fin area is
simply increased by an amount corresponding to 100% of the
immersed area) produced a very different pitching moment coef-
ficient from that of the undistorted configuration.

IV. Conclusions
Low-speed wind-tunnel tests were conducted to clarify the aft end
distortion effects on aerodynamic data for a model of an MK-82
bomb. To determine the relation between the increased base diameter
and the extended tail fin area of a distorted model, we performed
various tests based on DOE theory and processed data by using the
RSM. A regression model was found to successfully determine the
extended area of the tail fin needed to yield zero distortion effects on
Cm
. Our experimental data confirm that we need to increase the
exposed tail fin area by only 24% of the tail fin area that is immersed
due to the increased diameter of the base.
Downloaded by BEIHANG UNIVERSITY on July 1, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.40496

Fig. 7 Optimization plot for the response surface.

References
[1] Whitby, D. G., “Wind Tunnel Support System Effects on a Fighter
Aircraft Model at Mach Numbers from 0.6 to 2.0,” Arnold Engineering
Development Center, AEDC-TR-89-4, July 1989.
[2] Carman, J. B., Jr., “Store Separation Testing Techniques at the Arnold
Engineering Development Center, Volume I: An Overview,” Arnold
Engineering Development Center, AEDC-TR-79-1, Aug. 1980.
[3] Hill, D. W., Jr., “Investigation of Factors Affecting the Wind Tunnel
Measurement of Carriage-Position Airloads on Experimental Store
Models at Transonic Mach Numbers,” Arnold Engineering Develop-
ment Center, AEDC-TR-75-12, Feb. 1975.
[4] Montgomery, D. C., Design and Analysis of Experiments, 6th ed.,
Wiley, Hoboken, NJ, 2005, pp. 373–463.
[5] Myers, R. H., and Montgomery, D. C., Response Surface Methodology:
Process and Product Optimization Using Designed Experiments,
1st ed., Wiley, New York, 1995, pp. 1–123, 208–265.
[6] DeLoach, R., “Application of Modern Experiment Design to Wind
Tunnel Testing at NASA Langley Research Center,” AIAA Paper 98-
0713, Jan. 1998.
[7] DeLoach, R., “Tailoring Wind Tunnel Data Volume Requirements
Through the Formal Design Of Experiments,” AIAA Paper 98-2884,
June 1998.
[8] DeLoach, R., “Improved Quality in Aerospace Testing Through the
Modern Design Of Experiments,” AIAA Paper 2000-0825, Jan. 2000.
Fig. 8 Overlay plot for the response surfaces with a 1 prediction [9] DeLoach, R., and Cler, D. L., “Fractional Factorial Experiment Design
standard error SE . to Minimize Configuration Changes in Wind Tunnel Testing,” AIAA
Paper 2002-0746, Jan. 2002.
[10] DeLoach, R., “MDOE Perspectives on Wind Tunnel Testing
Objectives,” AIAA Paper 2002-2796, June 2002.
[11] Landman, D., Simpson, J., Hall, B., and Sumner, T., “Use of Designed
Experiments in Wind Tunnel Testing of Performance Automobiles,”
Society of Automotive Engineers, Paper 2002-01-3313, Dec. 2002.
[12] DeLoach, R., “The Modern Design of Experiments for Configuration
Aerodynamics: A Case Study,” AIAA Paper 2006-0923, Jan. 2006.
[13] Landman, D., Simpson, J., Mariani, R., Oritz, F., and Britcher, C.,
“Hybrid Design for Aircraft Wind-Tunnel Testing Using Response
Surface Methodologies,” Journal of Aircraft, Vol. 44, No. 4, July–
Aug. 2007, pp. 1214–1221.
doi:10.2514/1.25914
[14] Landman, D., Simpson, J., Vicroy, D., and Parker, P., “Response
Surface Methods for Efficient Complex Aircraft Configuration
Aerodynamic Characterization,” Journal of Aircraft, Vol. 44, No. 4,
July–Aug. 2007, pp. 1189–1195.
doi:10.2514/1.24810
[15] Barlow, J. B., Rae, W. H., Jr., and Pope, A., Low-Speed Wind Tunnel
Testing, 3rd ed., Wiley, New York, 1999, pp. 234–327.
[16] Chin, S. S., Missile Configuration Design, McGraw–Hill, New York,
1961, pp. 17–63, 94–106.
[17] Nesline, F. W., Jr., and Nesline, M. L., “Wing Size Versus Radome
Compensation in Aerodynamically Controlled Radar Homing Missile,”
AIAA Paper 1985-1869, Aug. 1985.
[18] Anon., Assessment of Experimental Uncertainty with Application to
Wind Tunnel Testing (S-017A-1999), AIAA Standard Series, AIAA,
Fig. 9 Extended area vs immersed area; velocity at 60 m=s. Reston, VA, 1999.
 598 OH AND PARK

[19] Coleman, H. W., and Steele, W. G., Jr., Experimentation and [22] Donegan, T. L., Bangasser, C. T., and Fox, J. H., “Computational Study
Uncertainty Analysis for Engineers, Wiley, New York, 1999, pp. 16– of Effects for the Joint Direct Attack Munition (JDAM),” AIAA
43. Paper 1998-2799, June 1998.
[20] Minitab, Statistical Software Package, Ver. 15.1.0.0, Minitab, Inc.,
State College, PA, 2006.
[21] Design-Expert, Statistical Software Package, Ver. 7.0.0, Stat-Ease, M. Miller
Inc., Minneapolis, MN, 2005. Associate Editor
Downloaded by BEIHANG UNIVERSITY on July 1, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.40496
 This article has been cited by:

1. Se-Yoon Oh, Seung-O Park, Seung-Ki Ahn. 2014. Experimental Investigations of Accuracy Improvement in Wind Tunnel
Testing Using Design of Experiments. Journal of the Korean Society for Aeronautical & Space Sciences 42:4, 291-297.
[Crossref]
2. Se-Yoon Oh, Jong-Geon Lee, Sung-Cheol Kim, Sangho Kim. Wind-Tunnel Testing of an Inertial Particle Separator Inlet
Using Design of Experiments . [Citation] [PDF] [PDF Plus]
3. Se-Yoon Oh, Seung-O Park, Seung-Ki Ahn. 2013. Experimental Investigations of Systematic Errors in Wind Tunnel
Testing Using Design of Experiments. Journal of the Korean Society for Aeronautical & Space Sciences 41:5, 335-341.
[Crossref]
Downloaded by BEIHANG UNIVERSITY on July 1, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.40496