Wind tunnel and test

techniques
S U B M I T T E D BY
ARJUN R KHEDEKAR
SUBMITTED TO
NIKIL MANE
 Abstract
layer wind tunnel, and the flow control strategies are applied to the sole upstream machine. The
experiment reveals a relationship, so far never observed experimentally, between the amount of yaw-
induced power losses and the implemented level of derating. Furthermore, combining derating and
yawing can only be beneficial, in term of improved power production at the cluster level , for modest
derating value.
The flap is deployed in one unique setting, and its cross section is morphed differently in take-off and
landing to get the necessary extra lift for the specific flight phase. Moreover, during the cruise, the tip of
the flap is deflected for load control and induced drag reduction. Before manufacturing the first flap
prototype, a high-speed (Ma = 0.3), large-scale test campaign (geometric scale factor 1:3) was deemed
necessary to validate the performance improvement brought by this novel system at the aircraft level.
As a proof-of-concept study, this paper mainly investigates the feasibility of a transonic flutter wing model
fabricated by the proposed methodology based on the AM technique and post-processing, which realizes
not only aerodynamic shapes but also designed elastic and modal characteristics, to evaluate typical
transonic aeroelastic responses such as the LCO and the transonic dip. The geometrical accuracy of
additively manufactured wing models is firstly accessed. Structural/aeroelastic characteristics of an
additively manufactured wing model for wind tunnel testing are then evaluated numerically and
experimentally.
 Abstract
Compared to traditional closed high-speed wind tunnels, this wind tunnel is endowed with remarkable
advantages of ample test chamber space, less interference from the tunnel wall, flexible model support
mode, and adjustable continuous variation of the Mach number.
In this paper, a flow field control system is designed for the 2 m HFWT by comprehensively using
advanced control technologies such as neural network, gain scheduling, feedforward control, and adaptive
control. Through practical application tests, it is proved that the proposed control system successfully
solves the problem of high-precision flow field control under continual depletion of storage tank pressure,
and realizes distinctive functions of adaptive static pressure matching and continuously varying Mach
number at supersonic speed.
The similarity law is verified by an HIFiRE-5 air vehicle, and the separation trajectories of wind tunnel tests
and the real air vehicle are obtained by numerical simulation. The research shows that the similarity law
derived in this paper can ensure that wind tunnel free-flight tests have the ability to predict the two-stage
separation characteristics of real parallel vehicles.
Aiming at the above problems, this paper innovatively proposes an aerodynamic intelligent identification
method, that is the transfer learning network based on adaptive Empirical Modal Decomposition (EMD)
and Soft Thresholding (TLN-AE&ST). Compared with the existing aerodynamic intelligent identification
model based on deep learning technology, this study introduces the transfer learning idea into the
aerodynamic intelligent identification model for the first time.
 Introduction
Innovations in current aviation have allowed for the design of more flexible aircraft structures enabling the
shape adaptation of the aerodynamic surfaces according to the demands of different flight phases, even at
very high speeds. These designs allow for the passive modification of the airframe configuration under the
external loads to get optimized performances for each flight condition (aeroelastic tailoring). The very same
task can be even more effectively carried out through the active and controlled morphing of the aircraft’s
aerodynamic surfaces, especially the wing, the control surfaces, and the flaps.
Finally, the geometrical precisions and structural/aeroelastic characteristics of the fabricated flutter wing
model have been evaluated through a series of experiments and observations. The characteristics of the AM-
based wing are also discussed by com-paring the experimental results and numerical solutions. The new
fabrication methodology of transonic flutter wing mode with the additive manufacturing technique achieved
the dimensional precision of additively manufactured flutter wing model being around less than ±0.16 mm,
and the surface roughness was about Ra 1.1μm. The present approach allows the cost-effective construction
of transonic flutter wing models and the investigation of tran-sonic aeroelastic phenomena such as the LCO
and transonic dip.
In contrast to traditional small-size closed wind tunnels, the 2 m HFWT not only has an apparent increase in
the volume of the settling chamber, but also greatly extends the distance between the flow field parameter
detection point and the corresponding control actuator, which makes the flow field characteristics of fast
time-varying and large lag more striking. Also, the unique structure results in an irregular pressure
fluctuation in test chamber, further aggravating the nonlinearity of the flow field.
 Introduction
According to the characteristics of parallel stage separation, compared with the previous static force
measurement and Captive Trajectory Simulation (CTS) test technology, the dynamic wind tunnel test
method has more advantages: the mass of the two-stage separator of parallel vehicles is closer, as is
the two-stage volume, the aerodynamic interference is serious, and the stage separation greatly
influences the two stage attitude. The former two test methods of static force measurement and CTS
test technology are not competent. To solve these problems, the multi-body separation free flight
wind tunnel test method is advantageous, which is very suitable for parallel stage separation
research.
A new intelligent aerodynamic force identification model based on deep learning is proposed in this
paper. The model is used to identify and filter out the inertial force signal and instrument noise signal
from the FMS output signal in the wind tunnel test, so as to obtain ‘‘pure” aerodynamic force. The
model will greatly reduce the design difficulty of strain gauge balance with high precision, low cost
and mature development, and improve the accuracy of pulse combustion wind tunnel aerodynamic
force test.
Compared with the existing aerodynamic intelligent identification model based on deep learning
technology, this study introduces the transfer learning idea into the aerodynamic intelligent
identification model for the first time. The TLN-AE&ST effectively alleviates the problem of scarcity of
training samples for intelligent models due to the high cost of wind tunnel tests.
Introduction
 Methodology
Clearly, everything comes at a price, and the advantages brought by morphing technologies are accompanied
by relevant penalties, such as an increase in structural weight and the need for higher actuation power. These
factors must be properly accounted for during the design process. It is, however, true that as the number of
potential benefits increases, the greater the penalties that can be accepted to get those benefits are.
At the end of the games, what matters is that the improvements brought by the technology well justify the
drawbacks associated with its implementation. When dealing with morphing structures implementation on
large aircraft applications, the demonstration of airworthiness poses relevant engineering challenges. The
efforts needed to address these challenges become worthwhile if the technology induced benefits are
significant at the aircraft level and, above all, in correspondence with multiple flight regimes.
The wing model parameters were first predefined, by considering the required aeroelastic characteristics and
design constraints (such as the size of the test section of a wind tunnel), to enable experimental observations
of transonic aeroelastic instabilities in the wind tunnel experiment. The aeroelastic characteristics were then
evaluated by numerical aeroelastic simulations.
If all requirements for the wing characteristics are satisfied, the pre-design of the wing was fixed. The
fabrication parameters involving the AM process and the post-processing were chosen. The influences of the
processes were then estimated based on our previously developed database or experimentally evaluated with
small prototypes. If the parameters of each process caused an unacceptable amount of deviations in the wing
characteristics, the corresponding parameter(s) was modified.
Methodology
 Result
To further illustrate the improvement of the jet flow field for employing the adaptive static
pressure matching control, the comparison of the contours of the vertical and horizontal Mach
number distribution in the flow field of the test chamber under the matching/nonmatching
control.
When the adaptive static pressure matching control is disabled, as depicted in the left side of
Fig. 12, the jet uniform region is limited to a diamond-shaped area with a length of only 1.8 m,
and the Root Mean Square (RMS) of the Mach number in this diamond-shaped region is 0.0039.
When the adaptive static pressure matching control is enabled, the size of the jet uniform
region, by contrast, is significantly increased and expands to a square space with a length of not
less than 3 m and a height of not less than 1.4 m as shown in the right side of Moreover, the
RMS of the Mach number in the original diamond-shaped region further decreases to 0.0036,
and even in the square jet uniform region, the RMS of the Mach number is only 0.0114.
 Result
The aeroelastic characteristics of the fabricated flutter wing model were evaluated by performing a transonic flutter wind
tunnel experiment. The wind tunnel test was conducted at Transonic Flutter Wind Tunnel in Japan Aerospace Exploration
Agency (JAXA). The wind tunnel was a blow-down type and had a 0.6 m ×0.6 m closed test section.
The experimental environment for the wind tunnel test. Aeroelastic vertical deflections of the flutter wing model were
measured at two marker locations on the upper surface by using laser displacement sensors (Keyence Corp.). The first
measurement was at a location of 75 and 12 mm from the wing root and the trailing edge, while the second location was
156.5 and 10.5 mm from the root and the trailing edge. Strain gage rosettes were also attached near the wing root on
the upper and lower surfaces to measure bending and torsional strains.
The separation characteristics of a wind tunnel test using CFD simulation. The two-stage separation is consistent with
the real air vehicle separation in the two stages can also be separated smoothly and safely. The separation also mainly
depends on the downward movement of the first stage, and the first stage has a larger angular movement. On the
whole, the wind tunnel test results obtained from the wind tunnel test designed by the design method in this paper are
consistent with the separation law of the real air vehicle at 30 km and have high accuracy. The separation characteristics
of a real air vehicle at 40 km simulated by CFD.
Similar to the 30 km separation characteristics in the two stages can be separated smoothly and safely. The separation
mainly depends on the downward movement of the first stage, and the first stage has a larger angular movement.
Compared with the 30 km separation in aerodynamic pressure at 40 km is lower, the overall separation time is longer,
which also conforms to the physical law. The separation characteristics of a wind tunnel test at 40 km simulated by CFD.
The two-stage separation is consistent with the separation of the real air vehicle. The two stages can also be separated
smoothly and safely. The separation mainly depends on the downward movement of the first stage, and the first stage
has a larger angular movement.
Result
Result
 Conclusion
Although aerodynamic identification results of cross-model size are promising, some limitations of this work
should be pointed out. First, in this study, the FMS involved in the test is single. The aerodynamic identification
network with good robustness is not considered (by using the dataset of the existing full-scale FMS) and used
to guide the design of new FMS.
Based on a limited number of wind tunnel calibration tests, a valve characteristic neural network that can
accurately describe the valve characteristic for arbitrary Mach numbers and arbitrary pressure ratios is
presented and applied to the closed-loop flow field control for the first time, which greatly improves the
accuracy of the flow field control.
A composite control strategy of open-loop feedforward compensation and closed-loop feedback tracking for
the stagnation pressure control is proposed, which successfully solves the problem of flow field high-precision
control under continual depletion of storage tank pressure. Moreover, by replacing the test Mach number with
the actual nozzle profile, this composite control enables the 2 m HFWT to carry out the supersonic
continuously varying.
Mach number test for the first time in China. In order to expand the size of the jet uniform region in the test
chamber and improve the uniformity of the jet flow field as much as possible, an adaptive static pressure
matching controller is designed for the 2 m HFWT. Under this three-loop controller, the static pressure in nozzle
outlet and the static pressure in test chamber can be adjusted to satisfy a given matching condition while
stabilizing the stagnation pressure within the predetermined stable boundary by regulating a single actuator of
the pressure regulator.
 Conclusion
The investigation of the geometrical precision for the AM-based wing models with different post-
processing allowances showed that the smaller allowance for the post-process could achieve smaller
deviations in the as-built models. On the other hand, a large deformation in a wing model may be induced
during the post-processing due to the small volume. Therefore, it is recommended to optimize the post-
process allowance to achieve the best accuracy in a final model by considering the trade-off between the
geometrical deviation in an as-built model and the induced deformation during the post-process.
With the evaluations of the geometries and surface roughness, it was confirmed that the additively
manufactured flutter wing model realized sufficient quality in the external/internal geometries and
smoothness for reliable transonic wind tunnel testing. In addition, a series of static load and vibration
tests with the fabricated flutter wing model showed that the wing model fabricated by AM could
accurately reproduce the structural characteristics of the designed model.

A brand new architecture was therefore developed for the flap model in order to meet the
scaled dimensions requirements. Then CFD analyses via Ansys Fluent are performed to
simulate the most relevant test cases that are supposed to be replicated in the wind tunnel.
18 test cases were simulated, resulting from the combination of three values of Mach, two
values of angle of attack, and three flap settings/shapes. Pressure distributions and velocity
speed vectors were collected and stored in a database for each case.
Conclusion
Conclusion