You are on page 1of 22

CJA 1435 No.

of Pages 22
21 November 2019
Chinese Journal of Aeronautics, (2019), xxx(xx): xxx–xxx
1

Chinese Society of Aeronautics and Astronautics


& Beihang University
Chinese Journal of Aeronautics
cja@buaa.edu.cn
www.sciencedirect.com

3 A comprehensive survey on the methods of angle


4 of attack measurement and estimation in UAVs
5 L. SANKARALINGAM, C. RAMPRASADH *

6 School of Mechanical Engineering, SASTRA Deemed University, Thanjavur 613401, India

7 Received 5 January 2019; revised 6 November 2019; accepted 6 November 2019


8

10 KEYWORDS Abstract Angle of Attack (AOA) is a crucial parameter which directly affects the aerodynamic
11
12 Angle of attack estimation; forces of an aircraft. The measurement of AOA is required to ensure a safe flight within its designed
13 Angle of attack measure- flight envelop. This paper intends to summarise a comprehensive survey on the measurement tech-
14 ment; niques and estimation methods for AOA, specifically in Unmanned Aerial Vehicle (UAV) applica-
15 Angle of attack vane; tions. In the case of UAVs, weight constraint plays a major role as far as sensor suites are
16 Five-hole probe; concerned. This results in selecting a suitable estimation method to extract AOA using the available
17 Multi-hole probe; data from the autopilot. The most feasible and widely employed AOA measurement technique is by
18 Multi-hole probe calibration; using the Multi-Hole Probes (MHPs). The MHP measures the AOA regarding the pressure varia-
19 Nine-hole probe; tions between the ports. Due to the importance of MHP in AOA measurement, the calibration
20 Seven-hole probe;
methods for the MHP are also included in this paper. This paper discusses the AOA measurement
21 Virtual angle of attack sensor
using virtual AOA sensors, their importance and the operation.
22 Ó 2019 Production and hosting by Elsevier Ltd. on behalf of Chinese Society of Aeronautics and
Astronautics. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

23 1. Introduction manoeuvring etc. To design such control laws and to have a 30


better knowledge about the aerodynamics of the aircraft, the 31

24 The Angle of Attack (AOA) is one of the critical parameters in AOA is an important parameter which directly contributes 32

25 a fixed-wing aircraft because all aerodynamic forces are func- to the Aerodynamic forces in such flight condition. If one 33

26 tions of AOA. The requirement of the knowledge on AOA has an idea of AOA, aircraft can be controlled effectively in 34

27 becomes more pronouncing when considering control of an many unconventional missions including flying at lower speeds 35

28 aircraft in highly unconventional missions such as flying and higher AOA. 36

29 through constrained space, high angle of attack flight, military Under stall condition, the knowledge about AOA is highly 37
essential, and stall recovery is possible only when the aircraft is 38
flying at higher altitudes. But in case of UAVs of mini or micro 39
* Corresponding author.
size, stall recovery is tough as they fly at lower altitudes. Aug- 40
E-mail address: crprasadh@mech.sastra.edu (C. RAMPRASADH). mented with the mission requirements, mini flyers cannot 41
Peer review under responsibility of Editorial Committee of CJA. afford to carry huge weight. This poses a great challenge for 42
the engineers to design the AOA measuring instrument to be 43
small in size and be less in weight. 44
Production and hosting by Elsevier

https://doi.org/10.1016/j.cja.2019.11.003
1000-9361 Ó 2019 Production and hosting by Elsevier Ltd. on behalf of Chinese Society of Aeronautics and Astronautics.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
2 L. SANKARALINGAM, C. RAMPRASADH

45 2. AOA measurement techniques as the best location for installing the sensors throughout the 83
subsonic, transonic and supersonic flows.1 84

46 William Gracey summarized the initial methods for measuring Hermann et al.2 proposed the idea of using a small tapped 85

47 AOA; the methods discussed were the use of a pivoted vane, airfoil to measure the AOA (Fig. 2). The study was conducted 86

48 differential pressure tube and null seeking pressure tube. The to evaluate the performance of the airfoil probe in low-speed 87

49 pivoted vane is a device which aligns itself with the direction dynamic pressure environments. From the theoretical evalua- 88

50 of airflow. The upwash effect, asymmetry of the vane and tion, the increment in differential pressure coefficients has been 89

51 bending of the boom due to air loads, have great influence shown as a factor of 2 for the airfoil surface. A small rectangu- 90

52 on the measured angle. The magnitude of the upwash effect lar wing with NACA0012 airfoil was taken for the wind tunnel 91

53 depends on the diameter of the boom support. The axially tests and it was found that it produces significant differential 92

54 mounted pivoted vane (Fig. 1(a)) is a conical body with a tri- pressure coefficients to obtain the AOA. These differential 93

55 angular vane to the rear portion of the cone. The maximum pressure coefficients were fed to an onboard computer to pro- 94

56 range of this type of vane is ± 15°. It was mainly developed duce digital AOA data. In this work, the effects of Reynolds 95

57 for measuring AOA in missiles. The transverse mounted vane number and Mach number were ignored because of low 96

58 (Fig. 1(b)) is attached over the shaft to the side of the boom speeds. The pressure sensed at the taps contains dynamic pres- 97

59 support. To measure the sideslip angle, another vane oriented sure and AOA. From the theoretical study, they have found 98

60 90° to the Alpha vane is used. This device also measures the that the maximum differentials between taps were obtained 99

61 Pitot and static pressures, and this type of vane is mainly when the taps are located at about 5 % chord position. Fur- 100

62 designed to mount on the fuselage nose boom during the flight ther, it is indicated that the variation of the differential pres- 101

63 test of research aircraft. The differential pressure tube is a sure coefficient with AOA is linear. Two independent sets of 102

64 device which consists of two orifices oriented at equal angles pressure measurements are required to extract the dynamic 103

65 on either side of the longitudinal axis of the tube. The AOA pressure information by the computer. Therefore a minimum 104

66 is measured by using the pressure difference between the two of three pressure taps should be used to extract the AOA from 105

67 orifices. The magnitude of the pressure difference depends on the air data. 106

68 the shape of the nose of the tube and the angular position of The wind tunnel tests were conducted using a well tapped 107

69 the orifices. half wing mounted on the floor of the test section. The pressure 108

70 The hemispherical and cone type nose shapes were anal- differentials for more than a dozen pairs of taps were tested to 109

71 ysed, and the greatest sensitivity of the tube to AOA was find the best pair. The first tap at the top surface is at 7.5% of 110

72 achieved by orienting the orifices about 90° apart on the hemi- chord; tap 2 is on top just aft of the leading edge, and tap 3 is 111

73 spherical nose and by using a 90° cone angle in the cone type on the lower surface at 5% chord which is found to be the best 112

74 nose. The null-seeking pressure sensor consists of a rotatable pair. The pressure calibration curves were approximated using 113

75 tube with two orifices at equal angles to the axis of the tube. the quadratic function Dp ¼ qðaa2 þ ba þ cÞ, (q-dynamic pres- 114

76 Pressure sensors were used to detect the pressure difference sure, Dp-pressure difference between the taps) which is used for 115

77 between the two orifices, and a mechanism was incorporated each pair of pressure taps to obtain the AOA using: 116
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
117
78 to rotate the tube to null pressure position. The pressure differ-
b1  bb2 b1  bb2 c1  bc2
79 ence between the two orifices was equalized by rotating the a¼ þ  ð1Þ
80 pressure tube. A potentiometer was used to measure the angu- 2ða1  ba2 Þ 2ða1  ba2 Þ 2ða1  ba2 Þ 119

81 lar position of the probe. A study on position errors is where b = Dp31/Dp23, and the subscripts 1 and 2 refer to pres- 120
82 reported, and the place ahead of the fuselage nose is concluded sure differentials Dc31 and Dp23. The test probe angles were set 121
manually, and it is compared with the measured value. From 122
the comparison, they establish that the measured values are 123
within 0.1° of the manual set values for the test range of 124
14° to 6° Similarly, for a range of 16° to +16°, it is 0.5°.2 125
Even a small AOA error will have a significant effect on the 126
calculated drag polar. A laser-based instrument was developed 127
by McDevitt et al. to measure the AOA of the models during 128
wind tunnel tests. In conventional mechanical systems, the 129
measurements exhibit a large number of errors due to vibra- 130
tion but in the optical method, an incident beam is passed nor- 131
mal to the model surface, and it is unaffected until the model is 132
tilted. The inclination of the model will refract the beam. This 133
displacement is proportional to the angle between the surface 134
and the incident beam. A photodetector is fixed behind the ref- 135
erence medium to measure the displacement. The optical AOA 136
measurement setup is shown in Fig. 3, reproduced from Ref. 3. 137
The position of the incident beam relative to the detector 138
may change due to the model vibration, deflections due to 139
dynamic pressure and temperature. These limitations are over- 140
come by using a spatial filter mounted ahead of the refractive 141

Fig. 1 Axially mounted vane and transversely mounted vane (1 1 medium along with the collimated light beam. There will 142

in = 0.0254 m). always be an incident ray at the same calibrated position rela- 143

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
Methods of angle of attack measurement and estimation in UAVs 3

Fig. 2 AOA sensor with a feedback system.2

pi ¼ qc ½cos2 ðhi Þ þ esin2 ðhi Þ þ p1 ð2Þ 170

where pi is pressure at the ith HI-FADS orifice, hi is a function 171


of the surface location and the effective angle of attack and 172
sideslip. e is high-angle of attack flush air data sensing model 173
calibration parameter. p1 is free stream static pressure. qc is 174
freestream compressible dynamic pressure) A least square 175
regression method was used to solve the equations to get the 176
calibration parameters, and from the calibration for 9 and 25 177
port configurations, it was reported that both produced nearly 178
identical results and concluded that the calibration is mostly 179
influenced by the aircraft configuration.4 180
Wong5 analysed three types of electrolytic tilt sensors to 181
measure the AOA of a model in wind tunnel tests. These tilt 182
sensors contain an electrolytic fluid inside a sealed glass vial 183
Fig. 3 Schematic of optical AOA instrument.3 which has three electrodes (Fig. 4). The motion of the elec- 184
trolytic fluid creates variable resistance between the elec- 185
trodes.Fig. 5. 186
144 tive to the photodetector. The detectors were capable of pro- When the AOA is zero, the tilt sensor will be in a balanced 187
145 viding readings at a 0.01° resolution over the 18° range and position. The resistance between the centre electrode and exter- 188
146 0.03 ° resolution over 44° range with the time-dependent out- nal electrodes is equal. When there is a change in AOA, the 189
147 put of 60 Hz which were sufficient to provide the accurate real- electrolytic fluid gets displaced and creates a resistance change 190
148 time AOA measurements.3 between the electrodes. This change in resistance is propor- 191
149 A flush air data system was developed by Whitmore, which tional to the tilt angle. By performing various tests, it has been 192
150 consists of a matrix of 25 pressure orifices arranged in concen- identified that the mode RG-37 electrolytic tilt sensor is found 193
151 tric circles on the nose of the vehicle. It is capable of measuring having the highest accuracy. Even though these electrolytic 194
152 the angle of attack, sideslip angle, Mach number and pressure sensors have performed better than the conventional servo 195
153 altitude. Traditional methods are not suitable to measure the accelerometer, they have some limitations due to the moving 196
154 AOA at high AOA flights. F-18 (high alpha research vehicle electrolytic fluid within the sensor. Upside down measurements 197
155 – HARV) was taken as the research vehicle. The HI-FADS are not possible. Also, the measuring range and sensitivity to 198
156 hardware, calibration and algorithm development were briefly the roll interaction (because of the moving electrolytic fluid) 199
157 discussed in this paper. The test was conducted for a Mach disqualify the tilt sensors as an AOA measuring instrument.5 200
158 number range of 0.15 to 1.20 and an AOA range of 8.0° to A new method was proposed by Sergio Callegari et al. using 201
159 55°. The HARV research measurement system includes the capacitive strip sensors for the measurement of AOA and air- 202
160 data from the 25 orifices. Also, the pitch, roll and yaw atti- speed. These sensors were directly applied over the wing 203
161 tudes, linear accelerations, angular velocities, and air data (Fig. 56) which eliminates the protruding effect by conven- 204
162 from the wing tip air data booms were digitally encoded and tional sensors. This strip sensor is specially formulated for 205
163 telemetered to the ground for post-flight analysis. The tests UAV applications. The accuracy requirement on the individ- 206
164 were conducted for all 25 orifices and a subset of 9 orifices. ual sensor can be lowered by using a redundant number of sen- 207
165 The pressure data were taken at 25 samples/s. sors and suitable signal processing techniques. By this way, the 208
166 The pressure at the ith orifice of the HI-FADS system was devices can be made inexpensive. The dynamic pressure distri- 209
167 given regarding flow incidence angle in the semi-empirical: bution over the wing was reconstructed by applying these pres- 210
168
sure strips which are capable of measuring pressure 211

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
4 L. SANKARALINGAM, C. RAMPRASADH

Fig. 4 Electrolytic tilt sensor and its electrical schematic diagram.5

Park et al.7 has developed a null type Alpha senor. For the 236
measurement of AOA in UAV, the operating range of the 237
UAV was taken as 10 km in altitude, 100 km/h to 400 km/h 238
in speed and 20 °C to 50 °C in temperature. The probe was 239
designed by utilising the combination of uniform flow and 240
doublet flow. The pressure holes were located on h1, h2 loca- 241
tions; therefore the difference in pressure is given by 242
243
p1  p2 ¼ 2q1 v21 ðsin2 h2  sin h1 Þ
2
ð3Þ 245

where h is the mean angle between wind and location points on 246
the circle. If h2 = h1, then the differential pressure is zero. This 247
idea was utilised to design the null type Alpha probe. The cone 248

6 angle plays an important role in the sensitivity and linearity 249


Fig. 5 Capacitive pressure sensors over wings.
between the differential pressure and AOA. The increase in 250
cone angle results in more sensitivity. However, the linearity 251
between differential pressure and AOA is getting decreased, 252
and the yaw angle influences the linearity. From the calibra- 253
tion, the cone angle of 45° is found to be having good perfor- 254
mance. The diameter of pressure holes is determined as 4 mm 255
from the wind tunnel tests with 72 km/h wind velocity. A 256

Fig. 6 Cross section of pressure sensing strip. 6 MEMS pressure sensor is used to measure the differential pres- 257
sure which is given to the microcontroller to process and calcu- 258
late the AOA. Various sensitivity analyses were done for the 259
212 differences. The pressure strip has three layers including a sensors, and the errors due to pressure and temperature varia- 260
213 floor, a spacer and a deformable thin polyimide top with many tions are taken for the calculation. A software filter was used 261
214 pressure-sensitive spots. Each spot is made of a chamber with to filter the noises in the measurements from the MEMS sen- 262
215 two electrodes, one is fixed at the bottom, and the other is sor. From the calibration, it has been concluded that the sen- 263
216 movable with the flexible upper layer. The same internal pres- sor does not respond to 6 Pa differential pressure and the 264
217 sure is maintained between all the chambers by connecting performance of the Alpha sensor depends on the sensor accu- 265
218 them with miniature pipes. racy and location of the mount.7 266
219 One of the spots has an air inlet, and it acts as the reference Xu and Wang8 proposed a new method to measure the 267
220 unit. Plate and membranes theory helped them to state the attack angle of the flying projectile. The coordinates of the fly- 268
221 proportionality between the differences in capacitance (among ing projectile were measured and utilised to obtain the attack 269
222 the electrodes of various chambers) and the pressure difference angle of the projectile. The intersection measurement principle 270
223 between the sensitive spots (Fig. 66). was used to measure the attitude of the projectile. 271
224 The thickness of the strip is below 1 mm, and the resolution This system includes two measurement units which consist 272
225 is about 5–10 Pa. The pressure differences between the sensi- of a linear array CCD camera and a theodolite. The two cam- 273
226 tive spots and the reference spot were collected as the vector eras will be kept looking upward at some angle. The visual axis 274

227 elements as Dp ¼ ½ðp1  p0 Þ; ðp2  p0 Þ; :::; ðpn1  p0 ÞT . If many of the two cameras intersect and forms a plane in the air. When 275

228 numbers of strips are used, then the vector becomes R = (Dpij) the projectile goes through this plane, the two cameras will 276

229 which is a function of airspeed and AOA. Some of the analyt- record the location information of the projectile. When the 277

230 ical methods are used to find the AOA and airspeed from the projectile crosses the plane in some angle makes different parts 278

231 relation. The maximum likelihood estimation was used to of the projectile pass through the plane in different points 279

232 solve the equation using the Probability Density Function (Fig. 7). In the analyses it has been reported that their method 280

233 (PDF). From the accuracy analysis, they have concluded that of combining co-ordinate measurement and attack angle mea- 281

234 sensor error is of ±0.4% on a range of ±1500 Pa with two surement is a feasible and simple method for various 282

235 sensor strips over the NACA0012 airfoil.6 applications.8Fig. 8. 283

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
Methods of angle of attack measurement and estimation in UAVs 5

matrix), where xi = [ai, bi]T contains the AOA and AOS esti- 305
mates, and b is a 6  1 mode independent matrix of aerody- 306
namic airframe parameters. 307
  308
aðVa Þ 1
¼ ðATi Ai Þ ATi bðVa Þ ð4Þ
bðVa Þ 310

Eq. (4) gives the least square solution of a and b. Because a 311
and b depend upon airspeed Va, the analytical expression for 312
Va was computed for each mode. 313

Fig. 7 Principle of attack angle measurement.8 The main advantage of Bayesian estimation over the LSQR 314
is that it considers the noise in the sensor measurements. The 315
normal distribution of a and b for each mode is given by 316

X ¼ R n þ X where R is obtained by Cholesky factorisation,
T T
317

and X is the mean values of the current estimates. The perfor- 318
mance of the algorithm was validated using HILS simulation. 319
It is concluded by claiming that their approaches require less 320
computational resources compared with the classical EKF 321
method. It makes it suitable for low-cost and low-power 322
UAV systems.9 323
Martos and Rogers10 have developed a low-cost instrument 324
system to measure the AOA and AOS. This AOA system 325
works based on the differential pressure value. It was found 326
that the usage of un-normalised differential pressure does 327
not provide adequate accuracy throughout the aircraft AOA 328
Fig. 8 Infrared sensor.11 range. The calibration curve is linear with enough accuracy 329
at high AOA regions except for the stall region; therefore, 330
the AOA system which uses ‘‘un-normalised differential pres- 331
sure” cannot be used to measure the AOA at cruise or low 332
AOA flights. It is suitable only for high AOA flights or as stall 333
warning devices. The pressure was measured by using a 334
MA4525 differential pressure sensor. The full-scale output is 335
1.0 psi with an accuracy of ±1/4% of full scale. Arduino 336
Uno was selected as the microcontroller because of its cost 337
and simplicity. Two DFRDAS systems were installed at 338
38.9% of the chord in the left and right wing respectively. 339
The measured differential pressures were post-processed to cal- 340
culate the AOA. From various configurations, it has been 341
Fig. 9 Three-hole probe for measurement of pitch angle.32 found that ppfwd
45
(pfwd is the total pressure and p45 is the pressure 342
measured from the pressure port located on the flat 45° sur- 343
face) provides adequate accuracy in the measurement of 344
284 A real-time air parameter estimation scheme was developed
AOA throughout the aircraft operating range.10 345
285 and implemented on a fixed-wing UAV by Mohammad Sha-
A new type of AOA sensor was developed by Sumantra 346
286 qura et al. Hardware in the loop simulation was used to anal-
Bhattacharya et al. by using a ball mouse encoder and two 347
287 yse the performance and efficiency of the algorithm. The
infrared (IR) modules. This method can be implemented only 348
288 Bayesian approach of estimation was compared with Extended
for the wind tunnel tests. This instrument consists of an enco- 349
289 Kalman Filter (EKF) estimation. The approach is based on
der wheel which rotates with the model, and it cuts the IR high 350
290 the decomposition of the nonlinear dynamical model of the
beam coming out of the IR transmitter. This reduces the inten- 351
291 UAV into a finite set of linear models. By a mode selection
sity of the light beam received at the receiver and the change in 352
292 algorithm, the correct mode is selected based on multiple
light intensity is used to measure the angle by which the wheel 353
293 stages of filtering. The linear least square method or Bayesian
is rotated. Two IR modules are used to identify the direction of 354
294 filtering is used for estimation. An analytical expression of the
rotation using the phase difference between them. The location 355
295 airspeed was developed for both of the approaches and the
of the two IR modules along with the encoder wheel is shown 356
296 estimated a, b is used to update the airspeed estimate. The
in Fig. 811. 357
297 LSM solves the finite number of linear least square problems
The IR-1 module leads the IR-2 for a clockwise rotation; 358
298 to estimate a and b. This method is faster than the Bayesian
the IR-1 lags the IR-2 for anti-clockwise rotation. 359
299 approach because LSM solution matrices are the only function
h ¼ h þ Dh; Dh ¼ ð/:b:dÞ, / = +1 for anticlockwise rotation, 360
300 of the dynamic model. In the Bayesian approach, the noise is
and / = 1 for clockwise rotation (Dh is the change in angle 361
301 taken from the Inertial Measurement Unit (IMU), and it pro-
and it is given by the product of /, c and d. Where b is the 362
302 duces more accurate results. Hybrid system non-linear modes
angular resolution and d is the net output from the total num- 363
303 were used to estimate a, b and Va. The six DOF system of
ber of comparators). The accuracy of 90% was achieved by 364
304 equations were written as Aixi = b (Ai – 6  2 mode dependent

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
6 L. SANKARALINGAM, C. RAMPRASADH

365 this method compared with the conventional AOA measure- with the body axis. The maximum error of 2.58° was found 427
366 ment methods.11 at zero AOA.13 428
367 A concept to estimate steady and turbulent wind velocities A data-driven fuzzy model was developed by Singh et al.14 429
368 and aerodynamic coefficients of a fixed-wing UAV is proposed to estimate the AOA using the recorded linear acceleration, 430
369 by Wenz et al.12, which can be utilised to calculate the AOA airspeed, pitch angle, and elevator deflection data. The Takagi 431
370 and airspeed. By this method, the need for UAV parameters Sugeno fuzzy logic is used to model the algorithm. For every 432
371 is avoided, and only the sensor information is required. A stan- input, the triangular type membership functions were gener- 433
372 dard sensor suite was utilised, which contains a Global Navi- ated. The grid-based ruspini type fuzzy partitions are used 434
373 gation Satellite System (GNSS), an IMU and a Pitot static for input spaces. The relationship between the measured 435
374 tube. These data are fused with kinematic, aerodynamic and parameters and the AOA are complex and non-linear in nat- 436
375 stochastic wind models in an EKF. ure. One of the best approaches to represent such complex 437
376
  and highly non-linear systems is the fuzzy logic based 438
wr
a ¼ tan1 ð5Þ Takagi-Sugeno model. It has two steps: the first step is to iden- 439
378 ur tify the structure and the second step is to estimate the param- 440
379
  eters. The fuzzy rules are extracted from the set of online 441
vr
b ¼ sin1 ð6Þ recorded input and output training data. The rules represent 442
381 Va the localised linear model for the specific combination of sub- 443
382 divided input variables. The antecedent and consequent 444
384 fz ¼ KV2a ðCL;0 þ aCL;a Þ ð7Þ parameters were tuned using the training data. To increase 445

385 Eqs. (5)–(7) give the expressions to calculate the AOA, AOS the accuracy, the use of big data modelling is suggested. 446

386 and acceleration in the z-direction. (a is angle of attack, b is A study was conducted by Kushner et al.15 about the mea- 447

387 side slip angle, The velocity vector of the aircraft in body frame surement of AOA using optical methods. These optical sys- 448

388 is given by (ur , vr , wr ), the total z acceleration decomposed in tems utilised photogrammetry principle to measure the AOA 449

389 the body z- axis is given by fz ). of the model kept in the wind tunnel. The model is painted 450

390 From the simulation results, it has been concluded that the with reflective paints, or some of the retroreflective targets 451

391 estimator is capable of estimating the AOA with Root Mean are marked over the model. Two or more special-purpose cam- 452

392 Square Error (RMSE) of 0.33° and the sideslip angle with an eras are used to observe the model attitude changes. The 453

393 RMSE of 3.21° and the airspeed with an RMSE of 0.23 m/s.12 retroreflective markers are used to enhance the brightness over 454

394 A flush air data system was developed by Vidya et al.13 to the object, which allows the cameras to distinguish the model 455

395 estimate the AOA, AOS, Mach number and other parameters. from the background. A magnetic position sensor can also 456

396 A nose cap was installed to get the surface pressure measure- be used to measure the orientation of the model. The strength 457

397 ment without affecting the airflow over the aircraft. The FADS and direction of the applied magnetic field can be measured, 458

398 system has been tested in a subsonic wind tunnel facility. The and the change in orientation of the model disturbs the mag- 459

399 tests were conducted by manually setting the model’s AOA netic field. This change in magnetic field can be used to mea- 460

400 and AOS, and then the differential pressure (MEMS-based) sure orientation. Their less accuracy and disturbances from 461

401 sensors were used to measure the differential pressure between the metal parts of the wind tunnel make this method unsuit- 462

402 the ports. This data was utilised to estimate the AOA, and it is able for AOA measurements. NASA has currently developed 463

403 compared with the model’s manual set AOA. The authors say a Model Deformation Measurement (MDM) and OAOA 464

404 that most of the FADS instruments use absolute pressure val- (Optical AOA) system which uses high-resolution cameras to 465

405 ues, but a new attempt is made to use the differential pressure capture the images of the test model. The model is painted with 466

406 values to estimate the AOA. A simple aerodynamic model was reflective dye and the taken images are processed to measure 467

407 derived to relate the air data states to the surface pressures. the AOA and deflection due to air loads. 468

408 pi ¼ qc ½cos2 hi þ esin2 hi  þ p1 , where pi is the incident port


409 pressure and hi is the flow incidence angle between the surface 3. AOA estimation methods 469

410 normal at the ith port and the velocity vector. The AOA and
411 SSA were computed using the surface pressure measurements Freeman16 developed a method to estimate the angle of attack 470
412
  by extensive modelling of stability derivatives. The authors uti- 471
1 A
ae ¼ tan1 ð8Þ lised only the data from accelerometers and the control surface 472
414 2 B deflections. The algorithm requires tail aerodynamics for cal- 473
415
   0   culation. To reduce the complexity of modelling, the parame- 474
1 1 1 A ters like pitch, roll, Mach number and G-loading limitations
be ¼ tan tan tan cosae ð9Þ 475
417 2 B0 were considered for estimation. This method entirely relied 476
on the aerodynamic stability derivatives and the control sur- 477
418 where ae is the effective angle of attack, be is the effective side
face positions to estimate the AOA. This algorithm was limited 478
419 slip angle, A0 ¼ Cik sin2 kj þ Cji sin2 kk þ Ckj sin2 ki , B0 ¼ Cik coskj
over a flight range with an accuracy of 0.5° of AOA. Popowski 479
420 sinkj sin/j þ Cji coskk sinkk sin/k þ Ckj cos ki sinki sin/i , the cone
presented the problems regarding the estimation of the angle 480
421 angle is given by k. The MEMS-based pressure sensor which of attack and the angle of sideslip on a flying object. Angle 481
422 has a full-scale output of 140 kPa is used to measure the differ- of attack and sideslip estimation strategies which are based 482
423 ential pressure. It has been reported that the error in the esti- on measurements of linear velocity aspects of an object with 483
424 mation is more at zero AOA; it is due to the relative angle the Earth’s coordinates and on attitude angles of the object 484
425 between the nose cap axis and the vehicle body axis. The error are presented. Both of these measurements originate from 485
426 is least at 15° AOA because the nose cap axis is not aligned

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
Methods of angle of attack measurement and estimation in UAVs 7

486 the inertial navigation system, and velocity measurement is perform the estimation. The in-flight estimator and post-flight 546
487 obtained from the satellite navigation system. The use of iner- estimators work with an accuracy of error less than 0.5°. 547
488 tial and satellite navigation for the estimation angle of attack A multi-stage fusion algorithm has been developed by 548
489 and sideslip angle is presented. Practical contrast of these esti- Ramprasadh to estimate the aerodynamic angle in a MAV. 549
490 mation methods has been conducted based totally on logged The MAV of low aspect ratio wings were taken as a research 550
491 parameters of a flight onboard a Mewa aircraft. The authors subject. A non-linearity in lift curve was reported. The estima- 551
492 have addressed the effect of the wind in the estimated values tion has three stages to get the estimated value of AOA and 552
493 of the attack and sideslip angles.17 AOS. Using a flight simulation program, the true states of 553
494 Olhausen18 developed an instrumentation package that the aircraft are calculated and used to estimate the Euler angles 554
495 makes use of an inertial navigation platform to obtain aerody- through EKF in the first stage. The state vector was taken as 555
496 namic and performance flight test data. YF-16 flight test pro- x = [V, a, b, p, q, r]T (x is the state vector, V is the airspeed, 556
497 gram used this package to obtain timely and accurate data. p is roll rate, q is pitch rate, r is yaw rate) and the measurement 557
498 The author discussed the theoretical basis and practical aspects vector was taken as y = [V, p, q, r, heu, /eu] T (y is the measure- 558
499 of using this Instrumentation Inertial Reference Set (IIRS) ment vector, h is the pitch angle, / is the roll angle). This 559
500 along with selected results. The author addressed the use of approach produced a bias in the estimation and to remove 560
501 IIRS to obtain flight path acceleration, a, b, normal load fac- the bias, the actual accelerometer outputs were used in the 561
502 tor, rate of climb, airspeed, take-off and landing velocities and place of estimated values. The NR solver is used to estimate 562
503 distances, position error calibration through the Mach jump the AOA and AOS in steady state and used as pseudo mea- 563
504 region, and wind information. surements in the EKF estimation. This changes the measure- 564
505 Thacker19 determined the position errors of the AOA sen- ment vector to y ¼ ½V aLPF bLPF p q r heu ueu Ax Ay Az (Ax is 565
506 sors on aircraft using state estimation with flight test data. acceleration is x direction, Ay is acceleration is y direction, 566
507 USAF Test Pilot School (TPS) aircraft is used to obtain flight Az is acceleration is z direction). The pseudo estimation is 567
508 test data, and the data was analyzed using Kalman filter. Air- given by 568
569
509 craft AOA position errors are caused by aerodynamic factors
aPseudo ¼ aNR þ ðTC1  Dt  qÞ ð10Þ 571
510 such as local flow and upwash. The author used state estima-
572
511 tion in Kalman filter to estimate the AOA data. The measured bPseudo ¼ bNR  ðTC2  Dt  r  cosðaPseudo ÞÞ ð11Þ 574
512 AOA data and estimated AOA data were compared to get the
513 position errors. The author concludes that his method was where aNR is the estimated angle of attack using Newton 575
514 accurate enough to identify a hysteresis error in the T-38A’s Raphson solver, TC is the tuning constant, Dt is the time step, 576
515 AOA sensor of ±0.5°. bNR is the estimated sideslip angle using Newton Raphson sol- 577
516 Zeis20 has developed the concepts for the AOA and AOS ver. The data of a MAV with 0.3 kg of mass and 0.3 m of wing- 578
517 estimation using inertial reference platform. These concepts span is used for the simulation. The elevator disturbance of 579
518 were tested in flight using NASA F-15A. The estimation of 0.05 rad for 0.2 s was given to the MAV. It has been reported 580
519 a, b were done in real time and also using the in-flight data that the error in prediction was reduced to less than 1° after 581
520 as post-flight estimation. The in-flight estimator was developed implementing the NR solver. The AOA and AOS have been 582
521 using linear accelerations and angular rates. A linear recursive estimated for various control inputs, and their respective errors 583
522 model was developed for post-flight estimation. The author were measured. By analysing various results, it was concluded 584
523 developed a new algorithm based on total lift and moment that the estimation of aerodynamic angles depends on the tun- 585
524 of the aircraft to estimate the AOA and AOS in real time. ing of the process noise covariance matrix of the EKF.21 586
525 The state-space method was used to estimate the aerodynamic Long an Song22 developed a multi-stage estimation method 587
526 angles using the post-flight data. For the post-flight estimation, to estimate the AOA and AOS. The data from various mod- 588
527 a linear recursive Kalman estimator with INS and CADS data ules are taken and combined to estimate the angles. The data 589
528 was utilised. were taken from ten separate modules for the estimation. 590
529 The AOA can be found as a function of The velocity components in the body axis are utilised to pre- 591
530 a ¼ FnðCLWB ; Ma; hÞ, CLWB is wing body coefficient of lift, dict the AOA and AOS. From the simulation, it has been con- 592
531 Ma is mach number, h is altitude. By using the assumption cluded that their method of using multi-stage fusion estimates 593
532 of no wind condition, the AOA can be estimated using data the AOA with an error of less than 0.1° and the AOS with an 594

533 from an inertial navigation system, a ¼ arctan VVBx Bz


(VBz is the error of less than 0.2°. It has been claimed that the method 595

534 velocity in z axis, VBx is the velocity in  axis), and the inverse works accurately even at high AOA regions. 596

535 of tangent of velocities in z- and x-axis gives the AOA. For the The estimation of AOA and AOS in a MAV under turbu- 597

536 post-flight estimation, the state-space model was taken as lent condition has been addressed by Ramprasadh and Arya.23 598
In a simulation environment, the turbulence was created using
^_ ¼ AX ^_ gives the complete state model,
599
537 X ^ þ BU;^ Z ^ ¼ HX, ^ X
a Dryden model. Initially, the MAV’s attitude was estimated, 600
538 ^ is perturbation value of state vector, Z
X ^ is perturbation value and it is utilised to estimate the AOA and AOS. The Newto- 601
539 of measurement vector. By considering the wind conditions, nian mechanics is used for modelling the simulation of 602

540 the estimator becomes aðkÞ ¼ arctan wu00 þwðkÞþuðkÞ


(aðkÞ is angle of MAV. The state and input vector was given by x = [V, a, b, 603

541 attack at time k, w0 and u0 are initial velocities in z and x direc- p, q, r, w, h, /, xe, ye, H]T, u = [de, da, dr, n]T. The components 604

542 tions. wðkÞ and uðkÞ are velocities in z and x directions at time of wind along the body axis of the aircraft are derived from the 605
wind profile as
k), bðkÞ ¼ arcsin v0 þvðkÞ
606
543 (bðkÞ is angle of sideslip at time k, v0 is
v
vw ¼ Vw cosðww  pÞsinw þ Vw sinðw  pÞcosw, ww ¼ 0, vw is 607
544 the initial velocity in y direction. vðkÞ is the velocity in y direc-
wind velocity component along y-axis, Vw is wind velocity, 608
545 tion at time k). A discrete Kalman filter program was used to
ww is wind direction, w is yaw angle, ww is wind velocity com- 609

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
8 L. SANKARALINGAM, C. RAMPRASADH

Table 1 Error in estimation with and without turbulence.


Error statistics Estimation with turbulence Estimation without turbulence
apseudo(°) bpseudo(°) aEKF(°) bEKF(°) apseudo(°) bpseudo(°) aEKF(°) bEKF(°)
Mean 1.6 0.15 1.52 0.03 0.44 0.15 0.35 0.08
SD 0.68 3.02 0.68 3.42 2.1 1.2 1.8 0.94
Peak 10.58 4.83 10.2 3.92 5.58 2.83 5.09 2.85

610 ponent along z-axis. The atmospheric turbulences were mod- was used to generate the correction to the angular velocity 661
611 elled using a shaping filter. The error in estimation with and from comparisons of body axis observations of the Earth’s 662
612 without turbulence is tabulated in Table 1. magnetic field and gravity. The summary of the attitude esti- 663
613 It has been concluded that the algorithm captures the tur- mator algorithm is given by x^ j ¼ ^cjk xk ; ^
yj ¼ ^cjk yk , x
^j is vector
614 bulence accurately, but the mean error is high with turbulence 664

615 when compared with the mean error in the estimation without estimated or computed value expressed in the j-frame, ^cjk is 665
616 turbulence. direction cosine matrix from k-frame to j-frame, xk is Vector
666
617 The robustness of the AOA estimation algorithm has been
actual value expressed in the k-frame. It has been concluded 667
618 tested by Ramprasadh and Arya.24 An EKF fusion algorithm
that the algorithm is simple and has low computational 668
619 was developed to estimate the AOA in Mini Aerial Vehicle
requirements with convergence guaranteed and the main dis- 669
620 (MAV). A Monte Carlo analysis was carried out over the algo-
advantage is to have lower accuracy than other methods. 670
621 rithm using randomly generated aircraft parameters. The sen-
Tapolcai et al.26 explored the non-linear method of analysis 671
622 sor data was simulated using a zero mean white Gaussian noise
for stalling and developed a non-linear index. The behaviour 672
623 with few true states. These simulated sensor data was utilised
of the aircraft cannot be predicted accurately using traditional 673
624 to estimate the AOA. The robustness of the estimation algo-
linear theory at high AOA regions. It has been claimed that 674
625 rithm with the changes in aircraft parameters due to manufac-
their non-linear approach can be used to expose the undetected 675
626 turing assembly errors has been analysed. The data for the
behaviour by the linear approach. The data for T-2C aircraft 676
627 simulation has been taken from a low aspect ratio MAV.
was taken for analysis. From the analysis, it has been con- 677
628 The aerodynamic equations of the MAV were computed using
cluded that the traditional linear approaches failed to indicate 678
629 the data. The flight simulation was carried out using a trim
the non-linear behaviour. The dynamic matrix A and the input 679
630 algorithm, which solves the aircraft equations of motion by
matrix B were obtained to derive the nonlinear index. The gen- 680
631 equating Vdot, adot and qdot equations to zero with the condi-
632 tions h = a, p = 0, q = 0, r = 0, b = 0. The trim values were eration of non-linearity is solely by the term @CZ@aða0 Þ, which 681

633 obtained by solving Vdot, adot and qdot equations. These trim changes concerning AOA. A subregion was created around 682

634 values were used in the flight simulation to generate the true each increment. The minimum value of 5% of the max AOA 683

635 states like V, p, q, r, Ax, Az. The error in finding the aerody- of the trim condition is given in the sub-region. The discretised 684

636 namic angles was found to be less than 1°. The aircraft param- values’ range is composed of 100 points from minimum to 685
1:834
637 eters measured from the sensors may not give the accurate maximum, and the process is given by Bj ¼ 5:730C md
, j = amin: 686
e
638 reading each time due to manufacturing errors and assembly astep: amax. These A and B matrices are non-dimensionalized 687
errors. The simulation has been carried out using the flight kBj Bi k
639
by vBj ¼ kB ik
. 688
640 parameters with a standard deviation ranging from 0.1% to
From the 100 values, the highest value represents the non- 689
641 10%. From the analysis, it has been concluded that a small
linearity of the aircraft under this trim condition within the 690
642 deviation of 0.1% in aircraft parameters affects the estimation
sub-region. The obtained non-linearity indexes were utilised 691
643 largely in the high AOA region.
for the analysis and it was concluded that their method of 692
644 de La Parra and Angel25 developed a navigation system for
using non-linear theory is capable of capturing the non- 693
645 SIVA family of UAVs. The authors used measurements from
linearity behaviour of the aircraft. 694
646 low cost sold state IMU, magneto-resistive magnetometer, sin-
Wang and Cai27 have presented a generalised 695
647 gle receiver inexpensive GPS, absolute and differential pressure
differentiation-integration observer to estimate the multiple 696
648 transducers. Three modules were used in the navigator, an atti-
integrals and high-order derivatives of a signal. The primary 697
649 tude estimator, a position and velocity estimator, and a mis-
purpose of the observer is to estimate the required states like 698
650 sion management module. ‘‘Lyapunov theory” was utilised
velocity, acceleration, etc. from different sensors. The first 699
651 by the attitude estimator algorithm which computes the angu-
observer was designed to estimate the velocity and acceleration 700
652 lar velocity corrections from Earth’s magnetic field and grav-
from the position using the data from GPS. The second obser- 701
653 ity. Three static Kalman filters were utilised by the position
ver was designed to estimate the attitude angle and angular 702
654 and velocity estimator. This corrects the integration of local
acceleration from the angular velocity data of IMU. The third 703
655 level components of acceleration with GPS, pressure derived
observer was designed to estimate the position and velocity 704
656 altitude and airspeed measurements. The mission management
from the acceleration. This observer was tested using the sim- 705
657 module manages the mission data, mission phases and ground
ulations, and it has been found that the observer succeeded in 706
658 control modes. The Complementary Kalman Filter (CKF) is
estimating the values of integrals and derivatives of the mea- 707
659 used instead of a linear Kalman filter to integrate the angular
sured signals. 708
660 velocity. In this non-linear observer, a non-linear control law

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
Methods of angle of attack measurement and estimation in UAVs 9

709 A real-time aerodynamic parameter estimation algorithm Johansen et al.31 proposed a method to estimate the AOA, 770
710 has been designed and implemented on a small remotely SSA and the velocity of a fixed-wing UAV using the kinematic 771
711 piloted aircraft by Andreas et al. The EKF was used for the equations. The aerodynamic models were not used for the esti- 772
712 parameter estimation. The noise covariance was used instead mation, but the estimator requires the airspeed and attitude of 773
713 of a forgetting factor for tuning. A small UAV was taken as the UAV. The required measurements were taken from the 774
714 the test aircraft. All the necessary data were taken using the standard sensors like GNSS, IMU and pitot-static system. A 775
715 appropriate sensors. The aerodynamic and propulsion data rotation matrix Rbn is used to transfer the North-East-Down 776
716 of that aircraft were computed using the wind tunnel tests. values to body fixed values. The AOA and SSA are given 777
717 An onboard computer was used to run the estimation algo- regarding the velocity vector as a ¼ tan1 ðwr =ur Þ, 778
718 rithm. The data required for the estimation were taken from b ¼ sin1 ðvr =Va Þ. The EKF has been used for the estimation, 779
719 the five-hole probe, infrared attitude sensors and INS. The and the experiment was done using three different UAVs. 780
720 Taylor series expansion was used to estimate the parameters The experimental vehicles were the NTNU Penguin, NASA 781
721 using the available data. A storage device is not used onboard. sierra, and NTNU X8 UAV. The validation of the estimated 782
722 Therefore the estimated parameter was transmitted to the values with the true values is not done; therefore the accuracy 783
723 ground station in real time. Three modes of flight test were car- of the estimators was not known. 784
724 ried out, manual, auto1 (Attitude control) and auto2 (way-
725 point/ manoeuvre navigation using the results from the 4. AOA measurement using multi-hole probes 785
726 estimation), and it has been concluded that this approach gave
727 good feasibility and good results with some minor limitations
The multi-hole probes have many advantages over other meth- 786
728 because of the non-identification of few derivatives.28
ods for measuring the AOA. Various measurements like total 787
729 Lichota and Lasek29 used the maximum likelihood estima-
pressure, static pressure, and flow velocity in three directions 788
730 tion output error method to estimate the AOA. The aircraft
can be measured using multi-hole probes. Their simplicity, 789
731 was modelled as a rigid body with oxyz as the body-fixed coor-
low cost and less maintenance are some of the added advan- 790
732 dinate system. The rotation matrix Kv was used for vector
tages over other methods for AOA measurements. 791
733 transformation from a body-fixed coordinate system to an
734 inertial coordinate system. For a short period motion, the
(1) Operating principle 792
735 AOA is given by w  au0 (w is the vertical velocity, u0 is the
793
736 trimmed state velocity in x direction), and when the turbulence
  The basic principle behind the operation of MHP is the 794
737 is included, the equation becomes a_ y ¼ Za a þ Zq þ u0 qþ
variation of pressure difference over the body when it is 795
738 ZdH dH þ wy (a_ y is the estimated angle of attack, Z is vertical immersed in the flow. Fig. 932 shows a pitch probe with pres- 796
739 force, q is pitch velocity, dH is elevator deflection angle, wy sure taps one, two and three at the angle h = 0°, 45° 797
740 is the process noise). The minimised error vector ‘‘v” for the and  45°. When the probe is kept parallel to the flow, the 798
741 MLE estimation based on the output equation and measure- stagnation pressure will be measured at h = 0°, and the other 799
742 ment ‘‘z” is given by vk ¼ zðtk Þ  yðtk Þ. The MLE is based two pressure tapping will measure the same pressure. When the 800
743 on maximising the conditional probability with the set of probe is kept at an angle, the pressure measurements include 801
744 unknown parameters H, which describes the object. The min- the velocity, the probe incidence angle, and the static and 802
745 imum Levenberg-Marquardt algorithm was used to find the dynamic pressure. 803
746 cost function. Small perturbations and central differences for- 804
q
747 mulae were used to get the gradient matrices. The Levenberg- p1 þ V21 ¼ pð45  aÞ þ 2qV21 sin2 ð45  aÞ ð12Þ 806
748 Marquardt algorithm has a limitation that it requires pre- 2
807
749 identification to obtain the initial values, and total least square q
750 is used to overcome that limitation Y þ e ¼ ðX þ lÞH (Y is p1 þ V21 ¼ pðaÞ þ 2qV21 sin2 ðaÞ ð13Þ 809
2
751 lateral force, X is longitudinal force, l is variables noise, H 810
q
752 is pitch angle). The measurements were sampled at a frequency p1 þ V21 ¼ pð45 þ aÞ þ 2qV21 sin2 ð45 þ aÞ ð14Þ 812
753 of 25 Hz. It has been concluded that a suitable method has to 2
754 be used to obtain the systematic error, which will reduce the Eqs. (12)–(14) are solved to get the probe incidence angle 813
755 bias in the AOA estimation. ‘‘a”. The number of holes at the tip of the probe varies as 814
756 Fekri and Mobed30 designed a controller using Linear per the range and accuracy of the measurement. The common 815
757 Quadratic Regulator (LQR) methods. The measured outputs Pitot-static probes can be used to measure the incidence angles 816
758 were used as feedback to the controller. Due to minimization of less than 10°; five-hole probe can measure the AOA up to 817
759 of the cost function, the controller had some issues with system 55°, seven-hole probe can measure the AOA up to 75°, and 818
760 input and states. Kalman filter is used to estimate the values twelve- or eighteen-hole probes are used to measure the inci- 819
761 from the noisy outputs. The estimation of AOA and theta dence angles larger than 75°. The analytical methods are not 820
762 has been done using linear Kalman filter and extended Kalman sufficient to find the accuracy of the Multi Hole Probe 821
763 filter. The simulation has been done with a biased white Gaus- (MHP). The experimental calibration of the MHP concerning 822
764 sian noise. The results of Linear Kalman Filter (LKF) and known incidence angle is used to find the accuracy of the 823
765 EKF were compared. From the simulation, it has been con- MHP.32 824
766 cluded that the EFK performs better than LFK. The EKF esti- CFR Nowack described a calibration method for a five- 825
767 mates the AOA with zero bias, but LKF estimation has a large hole spherical Pitot-tube. Two Cartesian angles were used to 826
768 value of bias in the estimation. The EKF is recommended over determine the direction of the velocity vector. The five-hole 827
769 the LKF for the estimation of AOA. probe was constructed with a diameter of 14.75 mm and the 828

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
10 L. SANKARALINGAM, C. RAMPRASADH

829 holes were of the diameter of 1 mm. One hole was positioned
830 at the middle, and the other four holes were positioned at
831 45° from the central hole. ‘‘van der Hegge Zijnen (1929)”,
832 one of the pioneers in the development of spherical Pitot
833 probe, found that the probe sensitivity is maximum when the
834 four holes lie between 40° and 45° to the central hole. The incli-
835 nation factors were formed using the pressure measured from
836 the holes. The manometer reading is taken as h and K is the
837 dimensionless factor obtained for a meridian plane with con-
838 stant angle w. The factors are dependent on head geometry
839 and flow directions. The calibration curves can be used to find
840 the flow angles within the range of 65 and +65.33
841 A new method of calibration for five-hole spherical and
842 hemispherical pitometers has been developed by Wright. The
843 relation between the flow parameters and the pressure distribu-
844 tion over the probe was obtained from potential flow theory.
845 The author gave more importance to measure the conical
Fig. 10 Calibration curves of angle-tube probe.35
846 and dihedral angle. To reduce the misinterpretation of the
847 data, the calibration curves were formed using new factors
848 K/, Kv and Kp.
849 The relation between dihedral angle and measured pressure
850 is given by tand ¼ ðp1  p3 Þ=ðp2  p4 Þ. From the calibration
851 curves, Eqs. (15)–(17) were formed and these equations were
852 solved to get the flow direction:
853
855 Kp ¼ 4:5expð0:3946/2 Þ  3:5 ð15Þ
856
858 Kv ¼ 1:0297 þ 0:0705/2 þ 0:0266/4 ð16Þ
859
861 / ¼ 0:8509K/ þ 0:3008K3/  0:0879K5/ ð17Þ
862 It has been reported that the instrument accuracy is fully
863 dependent upon the environment conditions and the hemi-
864 spherical probes have more advantage over spherical probes.34
865 Treaster and Yocum35 discussed the calibration and appli-
866 cation of FHP. The non-nulling method was used for calibra-
867 tion. The author has taken two types of probes, one is prism
868 probe, and the other is an angle tube. The prism probe was cal- Fig. 11 Calibration curves of prism probe.35
869 ibrated over the Reynolds number range of 2000–7000, and the
892
870 angle tube was calibrated over the Reynolds number range of p4  p7
Ch4 ¼ ð18Þ
871 20,000 in water flow and 8400 in air flow. The yaw-pitch and p4  p3 þp
2
5
894
872 pitch-yaw methods were used for calibration. In the non- 895
873 nulling technique, the flow angularity can be determined by p3  p5
C/4 ¼ ð19Þ
874 using the dimensionless pressure coefficients. From the calibra- p4  p3 þp
2
6
897
875 tion, it has been shown that the prism probe has a small range
876 of Cppitch values. The relations given in Eqs. (18) and (19) were used to cali- 898

877 Therefore it is more sensitive to the flow in the pitch plane. brate four different types of seven-hole probes and the calibra- 899

878 A differential pressure sensor is used to measure the pressure tion range of the seven-hole probes was successfully 900

879 from the five holes with a reference pressure Pref, DPi = (Pi - extended.36 901

880  Pref). Figs. 10 and 11 give the calibration curve of angle tube Ostowari and Wentz37 developed a new method for extend- 902

881 probe and prism probe. The AOA and SSA were calculated ing the measurement range of the five-hole probe up to flow 903

from the grid of Cpyaw versus Cppitch .35 angularities of 85°. The work presented is valid only for mea- 904
882
suring the pitch angles, and the yaw ports were nulled during 905
883 A non-nulling calibration method was described by Everett
the experimentation. The calibration coefficients were adjusted 906
884 et al. for the seven-hole cone probes. A method to increase the
by replacing the central port pressure with the upwind port 907
885 measuring range of flow angles up to 75° was developed. This
pressure and the stalled downwind port pressure was replaced 908
886 calibration is valid only for subsonic flows because the relation
with the central port pressure. From the results, it is evident 909
887 between pressures and Mach number was idealized for isen-
that the new method of calibration has expended the measure- 910
888 tropic flow. For high angles, the ports 3, 4, 5 and 7 were in
ment range of the five-hole probe. 911
889 the attached flow region. Ports 1, 2 and 6 were in separated
Kjelgaard38 developed a technique for the calibration of the 912
890 flow region. The relation between pressure measured by the
hemispherical tipped five-hole probe. The potential flow theory 913
891 reliable ports and the flow angles were given by

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
Methods of angle of attack measurement and estimation in UAVs 11

914 over a sphere was used to derive the equations which relate the culate pitch, yaw, total and total minus static pressure coeffi- 978
915 flow parameters to the pressure measured by the probe. The cients. The subset 1 indicates the central port, 2, 3 indicates 979
916 probe used for calibration has a diameter of 0.12 inch. The ports in the yaw plane, and 4, 5 indicates ports in the pitch 980
917 probe consists of five holes in the front portion facing the flow plane. A preliminary pitch angle is calculated using 981
918 and six interconnected static ports located at eight probe diam- p0 ¼ 9:8ðCPP þ 0:15Þ  poff ; poff ¼ 4:5 is the pitch angle offset 982
919 eters distance from the tip. The four holes in the front are kept resulting from the probe mount. The preliminary yaw angle 983
920 at an angle of 45° to the central hole. The central port was is calculated by interpolating p0 and Cpy. The miniature five- 984
921 numbered as 1; the static holes were numbered as 2, the b ports hole probe was compared with the Kiel probe to prove its 985
922 were numbered as 3, 4, and a ports were numbered as 5, 6. The qualification. At the Dean number of 119, a pair of counter- 986
923 potential flow theory can be applied for sphere only; therefore rotating Dean vortices was captured by the probe.40 987
924 theoretical calibration was generalised by including some of Deplot and Koschel41 developed a numerical approach 988
925 the experimental parameters. The FHP was mounted on a C- using a panel method to aid the design process of multi-hole 989
926 strut, and the data were obtained in 1° increment for eight dif- probes. The validation of the newly developed probes was 990
927 ferent velocities. Two sets of values were derived for the con- done using wind tunnel experiments; it is a time-consuming 991
928 stants h0, /0, A, B, C, D, E, F and G using the following process. To speed up the analysis, numerical methods were 992
929 equations: used to determine the pressure distribution on the probe head. 993
930
1=2 The panel method is based on the linearised potential equa- 994
ðDp2a þ Dp2b Þ sinð2hÞ tions which are written as an integral equation of a singularity
¼A ð20Þ 995
932 qp cosð2hÞ  cosð2h0 Þ distribution over the surface of a body. The model is discre- 996
933 tised by panels with specific singularity distribution. By evalu- 997
935 qp ¼ q½Cðcosð2hÞ  cosð2h0 ÞÞ þ Dðcosh  cosh0 Þ ð21Þ ating the strength of singularity of each panel, the local 998
936 velocity vectors can be determined, and it is used to obtain 999
938 p1  ps ¼ qðEcos2 h þ Fcosh þ GÞ ð22Þ the pressure coefficients. The pressure coefficients have been 1000
derived from the energy equation and these pressure coeffi- 1001
939 These constants were used to calculate the values of h and /
cients are used to compute the pressure distributions over 1002
940 which were converted into a and b values. The accuracy of the
the model. 1003
941 prone is within 0.5° for a and 1.5° for b, when the value of h is
Dominy Hodson42 have investigated the effects of Reynolds 1004
942 less than 50°.
number, Mach number and turbulence on the calibration of 1005
943 Zilliac39 described the non-nulling method of calibration
the five-hole probe. A ‘‘pseudo Mach number” has been intro- 1006
944 for conical seven-hole probes. A new analytical model was
duced for compressible flows to determine the flow angle and 1007
945 developed for the probe’s behaviour. The flow parameters were
other flow variables. The pseudo Mach number is given by 1008
946 obtained from the relation between the pressures measured by
k3 ¼ ppav1 . The probe was calibrated over the Reynolds number 1009
947 the probe. To increase the range of measurement, the probe
948 was chosen with a rake angle of 30° from the horizontal. A range of 7  103–8  103. From the analysis, it has been 1010

949 micro-stepping motor was used to position the probe in the reported that the sensitivity of yaw hole with Reynolds number 1011

950 known angles. The data were taken for various positions of happens in two regions with great influence. The observed first 1012

951 the probe. The AOA and AOS were given by region was at Reynolds number below 21  103 at large nega- 1013
952
  tive yaw angles. There was a low-pressure region at the tip of 1014
1 coshcoswsinc þ sinhcosc the probe due to the separation bubble. The second region
a ¼ sin ð23Þ 1015
954 cosb occurs at yaw angles of less than 8°. The pressure coefficient 1016
955 is intensive to yaw angles between 0° and +4° for Reynolds 1017
957 b ¼ sin1 ðcoshsinwÞ ð24Þ number greater than 50  103. The asymmetry virtually disap- 1018

958 A sectoring scheme was used to select the pressure values peared for Re < 20  103. Separated flow at low Reynolds 1019

959 only from the ports in the attached flow, during high flow numbers may lead to an error of 20 %, when the dynamic pres- 1020

960 angles. There were twelve pressure coefficients formed for the sure is measured at the region. Thus probes with 45° and 90° 1021

961 seven-hole probe using the measured pressures. For low flow cone angle were analyzed and it was reported that the sensitiv- 1022

962 angles, the flow is fully attached on the tip and three pressure ity of 45° probe is less than that of the 90° probe. The effects of 1023

963 coefficients were enough to form the alpha and beta coeffi- compressibility are less on the sensitivity of probes to Reynolds 1024

964 cients. These pressure coefficients were graphically plotted number changes. The effect of turbulence has been studied, 1025

965 between tangential and radial directions to relate them with and they found that the intensity of the turbulence affects 1026

966 AOA and AOS. The error in the method was within 0.5°. the probe calibration when Re < 20  103. Therefore the free 1027

967 A miniature five-hole probe with a tip diameter of 1.22 mm stream turbulence has to be considered during the calibration 1028

968 was developed by Ligrani to measure the three mean velocity of the probes. 1029

969 components in low-speed flows. The central tube is surrounded Hopper et al.43 developed four-hole probe with a frequency 1030

970 by four other tubes which have a taper angle of 45° with the response extended to 1.5 kHz and some techniques were devel- 1031

971 central tube. The probe was calibrated by positioning it man- oped to improve the frequency response and to use the probe 1032

972 ually using a traverse mechanism. The validyne pressure sen- in turbulent flows. The pressure tap holes are of 0.5 mm, 1033

973 sors were used, and the sensors are sensitive to the direction 0.25 mm and 0.3 mm diameter were constructed and cali- 1034

974 of gravity. The diaphragm deflection due to gravity is con- brated. The axial, tangential turbulence intensity was mea- 1035

975 tracted by multiplying a constant ‘‘k” with the sin of the pitch sured for 0.5 mm, 0.1 mm cobra probes and compared with 1036

976 angle and added to the pressure signal. The data were taken for hot wire anemometry. It has been concluded that the four- 1037

977 25 combinations of the angles, and these data are used to cal- hole cobra probe is capable of measuring all components of 1038

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
12 L. SANKARALINGAM, C. RAMPRASADH

1039 Reynolds stresses in developed turbulent pipe flow. The spatial the probe calibration flow. For conical heads, the velocity 1100
1040 resolution was improved for 0.5 tap diameter probe, particu- and pressure distributions change slightly with Reynolds num- 1101
 
1041 larly for the Reynolds shear stress component q u v when com- ber and Mach number at given yaw and pitch angles. The holes 1102
1042 pared with the 1.0 mm cobra probe. These results apply only to in the upwind have the smaller influence of Reynolds number 1103
1043 the turbulent core region and cannot be compared with the then the downwind holes due to flow separation. To overcome 1104
1044 data generated by the direct numerical simulation of the this, it has been suggested to use the seven-hole probes instead 1105
1045 Navier-Stokes equations in the wall region of the turbulent of FHP for measuring high angles in the flow. 1106
1046 flow. Babinsky et al.47 studied various geometries for the FHP 1107
1047 Fingersh et al.44 developed a technique to calibrate the FHP with planar sensor arrays. Due to unsteady and 3D flow fields, 1108
1048 for the wind turbine applications. The wind turbines operate in the frequency response of the probes should be high to get cor- 1109
1049 turbulent and unsteady environments which makes it difficult rect velocity measurements. For this, the pressure sensors were 1110
1050 for conventional methods to predict the flow properties. Previ- introduced in the head of the probe to get a high-frequency 1111
1051 ously, the flag vane instruments were used to measure the flow response (>100 kHz). To increase the sensitivity, various types 1112
1052 angles in the wind turbines, but the flag vane is accurate only in of fences were analysed, which are shown in Figs. 12 and 13 1113
1053 low-frequency measurements. The calibration has been done shows the pressure tapping locations of the five-hole probe. 1114
1054 for each roll angle, and the 3D plots of CPc , h, a as a function The probe head contains 41 pressure holes of 0.25 mm diame- 1115

1055 of DCPa and DCP0 were plotted. Various interpolation tech- ter and metal fences of 0.25 mm thickness. The Reynolds num- 1116

1056 niques were tried to define the surface constants but most of ber was varied between 1.5  104 and 4  104. Due to the 1117

1057 them failed, and finally, they have adopted a neural-network flexibility of the mount, the error of ±2° was introduced in roll 1118

1058 method with ten networks for each surface, which gave excel- angle and ±3° in pitch angle. 1119

1059 lent results. These curves were used to convert the FHP pres- Even though the accuracy is not enough for probe calibra- 1120
1060 sure data into qa, a, h. The dynamic analysis has been done tion, some of the conclusions were drawn over the use of 1121

1061 to check the sensitivity of the calibration curves to unsteady fences. The fences changed the flow patterns over the head, 1122

1062 flow conditions. Initially, there was a strong hysteresis with and from the analysis, it was found that the probe 3b gives 1123

1063 reduced frequency. It was eliminated by the phase shift due good results, and other open fence configurations have the 1124

1064 to pressure tube length. Due to wake shadow behind the tower, problem of vortices and flow separation. The sensitivity of 1125

1065 there is a dip in AOA at 180° which was not captured correctly the probe increases with the increase in fence height.47 1126

1066 by the flap probe due to poor frequency response, but the Five Robert Porro48 analysed various pressure probes to capture 1127

1067 Hole Probe (FHP) captures the dip in AOA because it is not the flow parameters in high-speed flows. The changes in flow 1128

1068 frequency limited. parameter occur on a millisecond time scale which requires 1129

1069 Morrison et al.45 developed a refined calibration technique high-frequency pressure transducers. The FHP used by the 1130

1070 to calibrate the FHP in the non-nulling mode. The 3D curve author had a 0.25 inch diameter, and the probe tips were opti- 1131

1071 fitting tool was used to get the 3D calibration surface equa- mised for supersonic flows. A miniature temperature- 1132

1072 tions. From the shape of the 3D surfaces, it can be determined compensated piezoresistive pressure transducer is used for 1133

1073 whether the probe can be used even with some abnormalities. the analysis which had a frequency response of 225 kHz. The 1134

1074 Using the calibration data, the author obtained the relation FHP with various geometrical configurations was tested at 1135

1075 between the pressure measured and the flow parameters. If Mach numbers of 1.6, 2.0, 2.5 and 2.8 at 2.5° increments in 1136

1076 any of the points is not correctly measured, it affects the accu- yaw and pitch directions. The ports two and four were aligned 1137

1077 racy in getting the flow angles. To overcome this limitation, a with the pitch axis. The ports three and five were aligned with 1138

1078 refined calibration technique was introduced, which reduces the yaw axis. The calibrations were done in the range of ±15°. 1139

1079 the effect of incorrect calibration points, and also it compen- Therefore, any result beyond this range is not valid. The pres- 1140

1080 sates the effects of non-symmetric probes. The arc curvature sure data were used to compute the pitch and yaw coefficients 1141

1081 increases with the increase in pressure, and from Pressure port which are used to get the flow angles. These probes were suc- 1142

1082 5 response to yaw and pitch it is identified that the probe is not cessfully tested in the wind tunnel to capture the flow field 1143
1083 aligned with the flow. The responses of each port to pitch and dynamics of a supersonic propulsion system on a 40 millisec- 1144

1084 yaw angles were plotted. If the plots were smooth and contin- ond time scale. 1145

1085 uous, the probe can perform accurately. If there are any abnor- Pisasale and Ahmed49 developed a method for extending 1146

1086 malities in the plot or the probe stops responding to changes of the calibration range of the FHP. Due to the occurrence of sin- 1147

1087 pitch and yaw angle, the probe should not be used in those gularity, the standard non-nulling calibration is limited to a 1148

1088 regions. The calibrated values were represented as 3D graphs certain range of flow angles. To non-dimensionalise the cali- 1149

1089 to evaluate the performance and the range of operation of bration coefficients, the difference between the total and static 1150

1090 the probes. pressure is used in the denominator. However, when the 1151

1091 Li46 studied the influence of Reynolds number on the probe denominator becomes zero which makes the value of pressure 1152

1092 measurements. The analysis was done numerically with the coefficients to infinity, various options were analysed to over- 1153

1093 conical seven-hole probe. The analysis was done for Mach come the singularity. The experiments were done in the wind 1154

1094 number of 0.3485 with Reynolds number of 5  104 tunnel with a constant velocity of 15 m/s. The initial try was 1155

1095 (Pref = 10.1013 MPa) and Mach number of 0.85 with Rey- to replace the denominator (p1  p) with the dynamic pressure 1156
1096 nolds number between 8.90  109 (Pref = 0.0709 MPa) and measured by the FHP. This eliminated the singularity, but the 1157
1097 6.42  105 (Pref = 0.5065 MPa). It has been concluded that CpYaw and yaw angles lose its one to one relationship, and for 1158
1098 the coefficients were influenced by the Reynolds number, espe- single CpYaw values, there were two yaw angles. Then the 1159

1099 cially when the measured flow Reynolds number differs from denominator p1  p was replaced with p1  pS (pS is static pres- 1160

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
Methods of angle of attack measurement and estimation in UAVs 13

1161 sure). This method avoided the singularity and also the multi- was found that the FHP probes are adequate to measure the 1221
1162 values, but it requires some reference data. Therefore, another three mean velocities, but recirculation is not present in the 1222
1163 approach was developed to overcome these limitations. Few measurement which makes it difficult to use FHP in the swir- 1223
 
p1 p
parameters were derived as Nx ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 p
ffi, Ny ¼ pDEN . ling glows.53 1224
2 2
1164 ðp2 p3 Þ þðp4 p5 Þ
1165 The DEN value is calculated, and using that DEN value, the
1166 CpYaw and CpPitch values were calculated using the relation
2 p3 1 pT
1167 CpYaw ¼ pDEN , CpPitch ¼ pDEN (pT is Total pressure). The error
1168 by this new approach was found within ±1.5%.
1169 Parameswaran and Jategaonkar 50 calibrated the Pitot-
1170 static systems and FHP using some of the data analysis tech-
1171 niques and flight test data. The parameter estimation tech-
1172 niques based on output error are used to estimate the AOA
1173 and SSA during the dynamic manoeuvres with rapid variations
1174 in the aircraft motion. To calibrate the flow variables, a com-
1175 plementary approach based on the Kalman filter technique is
1176 used. A classical approach is used for altitude determination
1177 through geometrical evaluation of photographs using tower
1178 flyby manoeuvres. It has been concluded that the estimated
1179 sensitivity factors for the flow angles measured by the FHP
1180 are reasonably accurate with manufacturer’s specifications
1181 and with the subject to corrections from misalignment and
1182 time delays caused by the recording equipment. Fig. 12 Types of fences.47
1183 Lee and Sang51 investigated the effects of Reynolds number
1184 on the non-nulling calibration of a typical cone type FHP. The
1185 analysis was done by changing the pitch and yaw angles from
1186 35° to 35° with the angle interval of 5°. The Reynolds num-
1187 ber is taken in the range between 6.60  103 and 3.17  104. It
1188 has been concluded that the Reynolds number significantly
1189 affects the pitch and yaw angle measurements. The static pres-
1190 sure coefficient is sensitive to the Reynolds number nearly all
1191 over the pitch and yaw angles range, and a remarkable Rey-
1192 nolds number effect was observed on the total pressure coeffi-
1193 cient when the absolute values of the pitch and yaw angles are
1194 larger than 20°.
1195 Zhai et al.52 developed a new method to evaluate the cali-
1196 bration data. The author made a precise relationship between
1197 flow quantities and measured pressures. A rake of 15 FHP was
1198 taken for analysis. Each probe has an outer diameter of
1199 2.5 mm. The calibration was done in the range of 34° to Fig. 13 Pressure tapping locations.47
1200 34° with intervals of 2° and at various Mach numbers of
1201 0.126, 0.142 and 0.157. From the standard method of calibra-
1202 tion, it was stated that, for the hemispherical probes, there is a
1203 vast non-linearity between the flow parameters and calibration
1204 coefficients. To overcome this limitation, a new set of calibra-
1205 tion coefficients were derived from the theoretical relations of
1206 the pressure distribution on the probe. The a and b values were
1207 computed using the new calibration coefficients. It has been
1208 concluded that, by introducing the new coefficients, the accu-
1209 racy of the measurement improved significantly and the work-
1210 load on the calibration reduced.
Fig. 14 Basic neuron architecture.54
1211 Caldeira-Pires et al.53 developed an FHP to measure the
1212 direction of mean velocity in turbulent flows. The calibration
1213 was done for ±25° pitch and yaw angles at 10 m/s velocity.
1214 The calibration was done using the standard method, and
1215 the calibration coefficients were found. The three-
1216 dimensional curve fitting was used to relate the AOA and
1217 SSA with the calibration coefficients. The values of X1 and
1218 X2 were calculated from X1 ¼ pp2 p  ; X2 ¼
p
3 p5 p4
p p
 . The FHP was
1 1
1219 experimentally tested to measure a tridimensional airflow at
1220 the exit of a premixing chamber. From the experiment, it
Fig. 15 Set of probes used for testing.57

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
14 L. SANKARALINGAM, C. RAMPRASADH

1225 Alberto et al.54 developed a non-intrusive multi-hole probe cessfully to obtain the flow angularity downstream of the 1267
1226 and a new approach for reconstruction of the air data param- rotor. 1268
1227 eters. The CFD analysis is used to determine the location of Kim et al.56 developed a five-hole multifunction probe and 1269
1228 holes for the pressure measurement. The neural network archi- calibrated it using multiple regressions. Two types of FHP 1270
1229 tecture was developed to reconstruct the air data from the data were developed: one is of cone head, and the other is of hemi- 1271
1230 obtained from the CFD analysis. Fig. 14 shows the basic neu- spherical head. The CFD analysis was used to optimise the 1272
1231 ron architecture. A weighted function (w) has been associated probe hole diameter and hole position angle. The optimal cone 1273
1232 with each input value (i). These weighted functions are multi- angle was found to be 70° with a 1 mm hole diameter. For 1274
1233 plied and summed together with a node bias value. The output hemispherical configuration, the optimal hole angle was found 1275
1234 was produced when the activation function (f) acts on the as 48° with a 1 mm hole diameter. From the CFD analysis, it 1276
1235 summed value (s). was found that the hemispherical probe gives better results. 1277
1236 The Artificial Neural Network (ANN) contains several lay- The hemispherical probe was manufactured and calibrated in 1278
1237 ers of nodes. A suggestion is made to increase the number of the wind tunnel. The a and b values were measured and using 1279
1238 holes to increase the redundancy of ANN configuration. The the above equations the values of h and / angles were 1280
1239 supervised learning method is used to make the ANN learn computed. 1281
1240 by giving the sample input–output data pairs. The ANN Pavel et al.57 integrated the combined AOA and AOS mea- 1282
1241 adjusts the weights until it produces the correct output for each surements into an existing air data system. The author manu- 1283
1242 input.54. factured twelve different types of probes and analysed their 1284
1243 Argüelles et al.55 developed a new method to calibrate a effect on the measurement of AOA and AOS. 1285
1244 Cylindrical Three-Hole Probe (CTHP) using the non-nulling Fig. 15 gives the set of probes taken for testing. The set A 1286
1245 method. New normalisation coefficients were developed to indicates the cobra probe (A) with 30° operating angle and the 1287
1246 prevent the singularity in the calibration coefficients: chisel’s probe (A) with 30°, 40°, 50°, 55°, 60°, 60° operating 1288

1247 Ca ¼ pL p
Q
R
, Cp0 ¼ p0 p
Q
C
, CpS ¼ p0 p
Q
S
, where angles. The set B indicates ball probes with 44° operating 1289

1248 Q ¼ pC  0:5ðpL  pR Þ, Ca is angular coefficient, pL is left hole angle. The set C indicates ball probe combined with AOA 1290

1249 pressure, pR is right hole pressure, Cp0 is total pressure coeffi- and AOS measurement with 100° and 44° operating angles. 1291

1250 cient, p0 is total pressure, pC is central hole pressure, pS is static The measurement devices are connected through the GPDB 1292

1251 pressure, Q is normalization coefficient. The normalised cali- bus to the DAQ system and MATLAB as shown in Fig. 16. 1293

1252 bration coefficients were obtained using the pressure distribu- From the experiments, it has been concluded that the probe 1294

1253 tion around the cylinder as pC ¼ pS þ pd Cp ðhÞ, size up to a diameter of 4 cm did not cause any problem. 1295

1254 pL ¼ pS þ pd Cp ðh  dÞ, pR ¼ pS þ pd Cp ðh þ dÞ(pd is dynamic The suggestion was made to choose bigger entry point holes 1296

1255 pressure, h is angle around the cylinder, d is construction angle and short pipes guiding to the sensor. For the aerospace appli- 1297

1256 of the probe). The value of Cp was determined using the poten- cations, the ball probes were suggested due to its output linear- 1298
ity. A suggestion was made to supplement the AOA and AOS 1299
1257 tial flow theory Cp ¼ 1  4sin2 h.The theoretical analysis was
system by a Pitot-static pressure input to simplify the calcula- 1300
1258 done with 45° and 60° angle spacing between the holes and
tions. To reduce the gravity influence over the sensors, they 1301
1259 it was found that the maximum pressure for left and right holes
should be placed in some particular orientation, which will 1302
1260 happens at ±45° and ±60° respectively. The limitation in
compensate the gravity.57 1303
1261 range occurs due to the unity value of the normalisation factor.
Argüelles et al.58 demonstrated the possibility of expanding 1304
1262 It occurs at the maximum hole angle. To overcome this limita-
the measuring range of the three-hole probe. In this study, the 1305
1263 tion, a new normalisation factor was developed and called as
influence of head geometry on the performance of three-hole 1306
pC þ pR  2pL pR > pL
QN ¼ . It has been concluded that probes was analysed. Fig. 17 shows the three different types 1307
1264 pC þ pL  2pR pL > pR
of head geometries taken for the analysis and the analysis 1308
1265 the new method of calibration increases the angular range was carried out for different hole separation angles. From 1309
1266 for both 45° and 60° probe. The probes have been tested suc-

Fig. 16 Measurement setup.57

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
Methods of angle of attack measurement and estimation in UAVs 15

Fig. 17 Different types of three-hole probes.58

1310 the analysis, it has been reported that the cobra type probes
1311 have the largest measurement ranges.
1312 The limit in the angular range is due to the double points
1313 emergence. For the cylindrical probe, the holes separation
1314 angle of 30° gives a higher angular range. The angular range
1315 is limited due to the duplicated zones. For the cobra probe,
1316 the hole separation angle of 45° produces higher angular
1317 range. The cylindrical probe has lower uncertainties to the flow
1318 variables, and it is more stable to Reynolds number variations
1319 and misalignments.58
1320 Akshoy et al.59 studied various algorithms for the calibra-
1321 tion of FHP and defined the non-dimensional pressure coeffi-
Fig. 18 AOA calculated using different numerical approaches.61
1322 cients in different ways. The five-hole probe used in this
1323 research is split into five zones, and for every zone, the pressure
1324 coefficients were formed using the influence of the central hole numerical modelling techniques. The calibration was done 1356
1325 (p5). The asymmetric nature in the sector map depicts the non- for the Mach number of 0.45–1.60. 1357
1326 symmetric nature of the FHP due to some manufacturing Fig. 18 shows the difference in calculating AOA using the 1358
1327 defects. It has been concluded that the new pressure normaliza- different numerical approaches. The sharp spikes in the trend 1359
1328 tion technique reported r2 value of 0.9979 for pitch angle and are due to water boiling which causes water percolating out 1360
1329 r2 value of 0.9910 for yaw angle, r2 value of 0.9761 for total and vapour pressure in the upper atmosphere.61 1361
1330 pressure and r2 value of 0.9938 for static pressure. When com- Crowley et al.62 studied the effect of turbulence on a multi- 1362
1331 pared to other normalization techniques, the new technique hole Pitot tube calibration. The hysteresis has been identified 1363
1332 gave the maximum r2 values.59 in a certain airspeed ranges due to flow instability. The hystere- 1364
1333 Lien and Ahmed60 studied the suitability of using MHP for sis caused an error of 30% in the calibration. The effects of 1365
1334 the measurement of skin friction in the turbulent flow. The introducing turbulence intensity were studied in this research 1366
1335 experiments were done in a pipe, on a flat plate and a swept work. The presence of hysteresis is stronger in the measured 1367
1336 forward facing step to simulate the one-, two- and three- values when the probe is oriented in the ranges of 20° ± 5° 1368
1337 dimensional turbulent flows. The non-dimensional pressure off the axis of the wind tunnel. A calibration factor Ci was 1369
1338 coefficients can be formed to include the wall effect which is DpIUT;i
defined to characterise the hysteresis Ci  Dpstandard , DpIUT the 1370
1339 given by
1340 pressure difference between the center hole and the ith off- 1371
p  p0
Cpn ¼ n ð25Þ axis hole of the Instrument Under Test (IUT), Dpstandard the dif- 1372
1342 0:5q1 U21 ferential pressure measured by a 1-D pitot tube. The flow tran- 1373

1343 When the probe is kept near the wall, the reading from the sition causes the hysteresis because of the recirculation zone. It 1374

1344 port which is near the wall gets affected. The values of pt and ps has been found that the hysteresis disappears by increasing the 1375

1345 were derived from the flow parameters. The relationship turbulence intensity. 1376

1346 between Dp* and sw* was given by It has been concluded that when the turbulence intensity is 1377
Dpd2 sw d 2 greater than 1%, the hysteresis is not present.62 1378
1347 lgðsw Þ ¼ MalgðDpÞ þ N, where Dp ¼ sw  ¼
4qv2
, (sw is
4qv2
Ravirai et al.63 designed a miniature four-hole probe with 1379
1348 wall shear stress, Dp is the difference between total and static the sensing area of 1.28 mm2 to reduce the measurement errors 1380
1349 pressure, d is the diameter of the probe). It has been reported due to large pressure and velocity gradients. The probe was 1381
1350 that the limitation of misalignment of the pressure method was specifically designed for high 3D turbomachinery flows. The 1382
1351 overcome by using the FHP. The probe should be small four-hole probe has been designed with a central hole, one 1383
1352 enough to be submerged in the boundary layer to get accurate yaw hole and two pitch holes, which differs from the conven- 1384
1353 calibration coefficients.60 tional four-hole probe. The central hole was chamfered at 90°, 1385
1354 Campbell and Brandon61 calibrated the Ares I-X five-hole and all other holes were chamfered at 35°. 1386
1355 probe using a combination of the wind tunnel, CFD and

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
16 L. SANKARALINGAM, C. RAMPRASADH

Fig. 19 Calibration curves of probe.63

1387 The calibration was held at the range of ±40°. Fig. 19 gives
1388 the calibration curves for the miniature four-hole probe. The
Fig. 20 Numerically generated calibration chart for
1389 change of probe pressure is so rapid at large yaw and pitch
Ma = 2.49.66
1390 angles, which implies that the calibration coefficients have
1391 higher sensitivity at large yaw and pitch angles. The accuracy
1392 increases with an increase in sensitivity but the operating range Navier-stokes (TLNS3D) code on the Cray-C90 platform. 1435
1393 decreases. The small chamfer angles result in lower sensitivity, Based on the accuracy and balance the one equation Spalart- 1436
1394 and large chamfer angles result in higher sensitivity. The Allmaras was used as the turbulence model. Fig. 20 shows 1437
1395 lookup table developed by Sitaram and Kumar was used to the calibration chart regarding non-dimensional pitch and roll 1438
1396 interpolate the flow angles with the measured pressures. It coefficients. From which the flow angularity can be deter- 1439
1397 has been found that the errors in yaw and pitch angles are mined. The results of CFD were validated using a series of 1440
1398 2.4° and 1.3°. It has been concluded that the four-hole probe windtunnel tests.66 1441
1399 is better than the five-hole probe to measure the 3D flows with Wildmann et al.67 used the Multi-Hole Probe (MHP) to 1442
1400 large pressure and velocity gradients.63 measure the turbulence of the airflow in flight up to 20 Hz. 1443
1401 Georgiou and Milidonis64 developed a subminiature FHP The remotely piloted aircraft MASC was used as the research 1444
1402 with embedded pressure sensors. For the turbomachinery aircraft. The meteorological wind vector ‘‘t” was calculated 1445
1403 applications, the probes should be as small as possible to get from navigation, flow and attitude measurements from the 1446
1404 highly accurate measurements. To avoid the oscillations of research aircraft. It is given as t ¼ tgs þ Mmf ðttas þ X  sp Þ(t 1447
1405 connecting flexible tubes, the pressure sensors were mounted meteorological wind vector, tgs ground-speed vector, ttas 1448
1406 at the end of the stem of the probe. The conventional non- true-airspeed vector, X vector of angular rotation rates, sp is 1449
1407 nulling method of calibration is used to calibrate the probe. lever-arm vector). The true airspeed of the aircraft was calcu- 1450
1408 The fourth-order polynomial regression method and the 3D lated using a and b values, which were derived from the pres- 1451
1409 curve fitting software are used to produce the relationship sure measurements of the MHP. 1452
1410 between the measured pressure values and the flow angles. Two methods for defining the dimensionless coefficients 1453
1411 The effectiveness of the probe is proved by measuring the flow were analysed and compared for the five-hole probes which 1454
1412 data of a reference turbine rotor. It has been found that the are given in Table 2. To capture the turbulence intensity in 1455
1413 small size probe measured the flow parameters with high the airflow, the signals were oversampled at 500 Hz. From 1456
1414 accuracy. the comparison, it was found that the MASC multi-hole probe 1457
1415 Lee et al.65 applied the Adaptive Network-based Fuzzy met the desired frequency response better than the M2AV 1458
1416 Inference System (ANFIS) method to calibrate the multi- system.67 1459
1417 hole probe in the non-nulling method. For the experiment, Pisasale and Ahmed68 used the potential flow theory to 1460
1418 the conical type FHP was used and the calibration process extend the calibration range of the FHP in low speed flows. 1461
1419 was automated using a data acquisition system. The ANFIS In the traditional method, singularity occurs when the denom- 1462
1420 system combined the benefits of reasoning capability of fuzzy inator equals zero. To overcome this, a coefficient Cpden is 1463
1421 logic and learning skill of a neural network. This system con- introduced and is given by Cpden ¼ Cp1  KCp p þ A. By the 1464
1422 sists of five layers which are fuzzy, product, normalised, de-
calibration, it has been found that for the value of K = 0.2 1465
1423 fuzzy and total output. It constructs the relationship between
and A = 1.05, the calibration range has been increased up 1466
1424 the input and output of each layer. It has been concluded that
to ± 80°. 1467
1425 the ANFIS method is efficient because the training error is less
The modified calibration coefficients were given by 1468
1426 than 7.2  103 and it has a high consistency of 0.0068 after 1469
1427 iteration. This demonstrates that this combination produced Cpyaw ¼ 2sinð2cÞtanb ð26Þ 1471

1428 better results than the conventional methods. 1472


1429 Milanovic and Kalkhoran66 developed a miniature conical Cppitch ¼ 2sinð2cÞtana ð27Þ 1474

1430 FHP of 30° half angle, and it was numerically calibrated at It has been claimed that even though the error percentage is 1475
1431 supersonic Mach numbers. the same for older and new methods, still the new methods 1476
1432 The spherical coordinate system is used to express the were favourable due to the avoidance of the singularity.68 1477
1433 quantities. The numerical calibration was performed using Jadhav et al.69 fabricated a five-hole probe and calibrated 1478
1434 the multi-block version of three-dimensional Thin Layers the FHP using non-nulling calibration method for a known 1479

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
Methods of angle of attack measurement and estimation in UAVs 17

Table 2 Two methods for defining dimensionless coefficients.67


Parameter Bohn et al. (1975) (M2AV) Treaster and Yocum (1979) (MASC)
 2 12 h
Dp P4
i dp1 þdp2 þdp3 þdp4
P4 P
1
5 i¼0 pi  5
1
j¼0 pj þ p0  14 4i¼1 pi 4

ka dp01 dp03 dp1 dp3


Dp dp0 Dp
kb dp02 dp04 dp2 dp4
Dp dp0 Dp
kq qdp0s dp0 q
Dp dp0 Dp
kp ps þdp0s p ps þDpp
Dp dp0 Dp

1480 flow field. The outside ports were arranged at 45° to the central Ca2  Ca3
Ca ¼ Ca1 þ ð30Þ 1529
1481 port. The calibration took place at the flow velocity of 25 m/s. 2
1482 It has been concluded that the fabricated probe is best suited 1530
1
1483 for the measurements of pitch and yaw angle of ±20°. Berez- Cb ¼ pffiffiffi ðCa2 þ Ca3 Þ ð31Þ
3 1532
1484 nai and Mlynár70 designed a spherical five-hole probe and cal-
1485 ibrated it at different angles of attack. The calibration The artificial neural network algorithm is used to improve 1533

1486
coefficients were given as kw ¼ p2 p w2
1
, kb ¼ pp2 p
p
1
, kp ¼ p2 p
w2
s
, the accuracy of the prediction. The final network had four hid- 1534
q2 2 3 q 2
1487 akor ¼ fðbÞ. The correction coefficient akor is measured when den layers with verified linear unit activation functions. It has 1535

1488 the probe is kept at p4 = p5. The dependency of the calibration been claimed that the prediction accuracy has increased up to 1536

1489 coefficients on the b angle was analysed and it was concluded 60% using the neural network.73 1537

1490 that this method of calibration can be used only where it is Benjamin and Thomas74 developed an open-access five-hole 1538

1491 possible to turn the probe to make p4 = p5. probe named oxford probe for aerodynamic measurements. 1539

1492 Ratneswar et al.71 developed an algorithm to calibrate the The probe geometry and drawing have been made open access 1540

1493 FHP in non-nulling technique and compared it with the null so that the end user can manufacture the Oxford probe using 1541

1494 technique. The calibration has been done by splitting it into the additive manufacturing methods. The probe was calibrated 1542

1495 five sectors. From the calibration constants, the value of a, using standard non-nulling calibration method. The flow 1543

1496 b, Cps, Cpt were interpolated. The least square curve fitting angles can be calculated with the calibration maps by using 1544

1497 method with matrix operation was used to obtain the calibra- the available pressure information. A 5 mm diameter probe 1545

1498 tion coefficients. From the comparison with the null method, it would produce 2.3° roll offset due to 0.1 mm misalignment 1546

1499 has been concluded that the null method of calibration is sim- from the nominal value. The misalignments can be corrected 1547

1500 ple but time-consuming, and for large flow angle measure- using the following roll correction equations: 1548
T cosbT sinaT sinhT 1 sinhT sinbT þcosbT sinaT coshT
1501 ments, a suggestion was made to use the non-nulling b ¼ tan1 coshT sinbcosa T cosbT
, a ¼ tan cosaT cosbT
. 1549
1502 technique. A total of 16 probes of various size and material were manu- 1550
1503 Shevchenko and Shmakov72 developed an analytical model factured and calibrated at a single Mach number of 1551
1504 to measure the flow angularity in subsonic and supersonic Ma = 0.3. All the calibration maps which were formed for 1552
1505 flows using a nine-hole probe. The pressure distribution over the probes were made open access to the public. 1553
1506 the sphere for various Mach numbers both in subsonic and Daniel et al.75 investigated the flow around the Pitot tube 1554
1507 supersonic regions was analysed. It has been found that the using the computational fluid dynamics to find the location 1555
1508 pressure distribution remains the same for various Mach num- of flow separation, installation effects and the differential pres- 1556
1509 bers. By using this results, the analytical model was formed, sure between the ports of annubar. The flow meter is used to 1557
1510 which can be used both in subsonic and supersonic regions. measure the differential pressure between the downstream 1558
1511
1   1   and upstream ports. Due to the suction effect, the measured 1559
Cb ¼ Cp7  Cp3 þ pffiffiffi Cp6  Cp2 þ pffiffiffi Cp8  Cp4 ð28Þ value needs to be corrected using a flow coefficient ‘‘K”, 1560
1513 2 2
1514
K ¼ 2 pqmnffiffiffiffiffiffiffiffiffiffi (qmn is dynamic pressure, Dpn is differential
pR 2Dpn qn 1561
1   1   pressure, qn is fluid density). Five turbulence models were anal-
Ca ¼ Cp5  Cp1 þ pffiffiffi Cp6  Cp2  pffiffiffi Cp8  Cp4 ð29Þ 1562
1516 2 2 ysed and compared with the experimental data to get the 1563

1517 The relation between the flow angularity and pressure data appropriate turbulence model. It has been found that the k- 1564

1518 is given by Eqs. (28) and (29) (Cb, Ca), and from the error anal- x model’s prediction is nearer to the experimental value. The 1565

1519 ysis, they have concluded that the results were in the accept- CFD analysis with various turbulence intensity and various 1566

1520 able range. shapes of the Pitot tubes was carried out using the k-x turbu- 1567

1521 Conlon et al.73 developed a new calibration method to lence model. From the various turbulence intensity analysis, it 1568

1522 reduce the calibration time. The seven-hole probe was taken was shown that the pressure distribution around the tube is not 1569

1523 for the analysis. The relation between the calibration coeffi- affected by the turbulence intensity. It has been concluded that 1570

1524 cients and the pressure measurements was given by the probe shape in the upstream has fewer effects in the calcu- 1571
p1
Ca1 ¼ pp4p
 , Ca2 ¼
p3 p6
 , Ca3 ¼
p2 p5
 . These three equations
lation when compared with the probe shape in the down- 1572
1525 p p p p
7 S 7 S 7 S stream. The largest change in the flow coefficient ‘‘K” 1573
1526 Ca1, Ca2, Ca3 are combined to form: appears at a distance of L/D < 11.75 1574
1527

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
18 L. SANKARALINGAM, C. RAMPRASADH

causes that incorrect non-dimensional pressure coefficients 1599


has been addressed. The new technique is an iterative process 1600
which takes the initial guess of the static pressure to start the 1601
iteration. The computed calibration coefficients were com- 1602
pared with the values in the database to get the flow angles. 1603
A regression coefficient was used to enhance the comparison. 1604
The curve fitting method is used for half of the domain to 1605
get the flow angles when one of the holes is blocked or faulty. 1606
It has been claimed that their technique predicts the flow angle 1607
with a maximum error of 0.2° and the algorithm automatically 1608
detects the faulty port and regenerates the database with fewer 1609
pressure ports.77 1610
Paul et al.78 designed a calibration rig for the velocity probe 1611
with a flow channel, probe manipulator and an axial fan drive 1612
system. The LabVIEW software was used to give the user 1613
inputs to the rotary table. From the experiments, it was found 1614
that the probe manipulators can be used for the measurement 1615
of pitch angle in the range of ±79° and yaw angle in the range 1616
of ±85° with 1/10° turning accuracy. Nekkanti and Kancherla 1617
79 analysed the effect of the chamfer angle of the FHP on the 1618
Fig. 21 Cobra probe head and different zones.76 calibration curves. The chamfer angle has been varied from 30° 1619
to 60° in 5° steps. The calibration was done in the range of 1620
30° to + 30° in 5° interval. For the analysis, a 9.6 mm diam- 1621
1575 Benjamin76 analysed the accuracy of a four-hole cobra
eter FHP was fabricated with a truncated cone tip, and the cal- 1622
1576 probe using Turbulent Flow Instrumentation (TFI) and devel-
ibration was done at a velocity of 25 m/s. From the analysis, it 1623
1577 oped a MATLAB Graphical User Interface (GUI) to analyse
has been concluded that the value of calibration coefficients 1624
1578 the data taken using the cobra probe. Fig. 21 shows the
increases with the increase in chamfer angle and this trend is 1625
1579 four-hole cobra probe and various zones of the four-hole
observed up to the chamfer angle of 50°. For the 50° chamfer 1626
1580 probe. The data were recorded for pitch angles from 45° to
angle, the calibration coefficients show maximum sensitivity, 1627
1581 45° and yaw angles from 45° to 0°.
but the chamfer angle of 30° has maximum range. 1628
1582 From the error analysis, it was found that the maximum
Bartosz et al.80 developed an integrated airflow probe with 1629
1583 error of 20.6% occurs at 30° yaw and 15° pitch at 89 m/s.
flow angle vanes specifically for the UAVs. The design process 1630
1584 Four positions were found where the cobra probe is not able
includes three stages: first is to formulate the size of the probe 1631
1585 to measure the values. From the uncertainty analysis, it was
and other sensors, the second stage is to select the appropriate 1632
1586 found that the uncertainty of the pitch and yaw angles to be
sensors and electrodes, and the final stage is to formulate the 1633
1587 ±0.5° and recorded velocity is within ±0.30 m/s to the actual
software algorithm and implementation. From the literature 1634
1588 value. The tests were conducted in the designed region and
and availability of brass tubes, the probe of 13 mm outer diam- 1635
1589 concluded that the cobra probe is not accurate for velocity
eter and 12.1 mm of inner diameter has been manufactured. 1636
1590 measurements in 17.6% of the designed region, for pitch and
The pylon for the AOA vane was placed at 175 mm from the 1637
1591 yaw measurements 20.9% and 12.1% respectively.76
tip of the probe and the second pylon for the side slip vane 1638
1592 Tolga and Guillermo77 developed a technique to process
was placed at 225 mm from the tip and built with 30 mm long 1639
1593 the data recorded from an MHP, which is more accurate than
brass tubes. For the electronic circuit, a specially shaped PCB 1640
1594 the conventional methods. Databases of non-dimensional
was created to fit in the probe, and the magnetic encodes 1641
1595 pressures were utilised instead of using the differential pressure
AS5046 were used to measure the angular displacement. The 1642
1596 levels.
total assembly was mounted on the UAV to get the in-flight 1643
1597 Fig. 22 shows the geometry of the FHP that was used in this
recorded data. A sharp increase of AOA is observed with a 1644
1598 research work and also the problem of hole blockage, which

Fig. 22 Geometry of FHP.77

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
Methods of angle of attack measurement and estimation in UAVs 19

1645 decrease in airspeed, and it has been claimed that the probe’s box neural network are used to increase the accuracy of the 1691
1646 measurements are as per the theory. virtual sensor. The AOA virtual sensor consists of Linear 1692
1647 Fatma and Schiffmann81 analysed the behaviour of the Parameter Varying (LPV) model and a neural network model 1693
1648 pressure probe in periodic flows at a resonating frequency. A a ¼ aLPV þ aNN ). The neural network is a black box model,
(^ 1694
1649 single-hole probe has been designed for measuring the steady and the LPV model is a white box model. The LPV 1695
1650 and periodic flows at high fluctuation amplitudes. This pro- (aLPV ¼ a0 þ aSP ) has the trimmed AOA (a0 ) and a linear short 1696
1651 duces a non-linear behaviour of the probe which is not able period approximation of AOA (aSP ). The trimmed AOA has 1697
1652 to be estimated using available models. The transfer functions been estimated using the T-S fuzzy model which takes the 1698
1653 were introduced to each harmonic to reconstruct the estimated Mach number, dynamic pressure, bank angle, position of cen- 1699
1654 data from the models. Three miniature single-hole probes with tre of gravity along with the X-axis and the aircraft weight. 1700
1655 the same inner and outer diameter were taken with different The short period AOA (aSP ) was estimated using the approxi- 1701
1656 lengths for the analysis. The siren disk was used to conduct mation shown below 1702
1703
1657 the dynamic response tests which can generate fluctuations  
b^1 r þ b^0 g
1658 up to 10 kHz. The modified Helmholtz resonator model was aSP ðkÞ ¼ q tan/sin/ ð33Þ
1659 used for estimation and the model under-estimates the flow r þ a^0 ðXðkÞÞ VT 1705

1660 compressibility and viscous force which creates an error in The structure of the virtual AOA sensor is given in 1706
1661 the prediction. A transfer function based on the Fourier series 1707

1662 was used to reconstruct the pressure data measured by the ^


aðkÞ ¼ a0 ðXðkÞÞ þ Hðr; a^0 ðXðkÞÞÞq/ ðkÞ þ aNN ðXNN ðkÞÞ ð34Þ 1709

1663 probe. The transfer function prediction is above 75% as the where aSP short period approximation of the angle of attack, r 1710
1664 reference, up to twice the probe resonance frequency and it is the forward shift operator, a^0 ðXðkÞÞ is the TS fuzzy model, 1711
1665 drops to 40% when the excitation is higher than twice the res- g
tan/sin/ describes the approximation of the output of the 1712
VT
1666 onance frequency. pitch rate sensor during turns with constant pitch rate, con- 1713
stant bank angle and constant turn rate, ^ aðkÞ is the structure 1714
1667 5. Virtual AOA sensor of the entire virtual AOA sensor, aNN denotes the output of 1715
the neural network model. 1716
1668 The virtual sensors are the software-based devices which use The virtual sensor has been validated by validating LPV 1717
1669 the measurable signals to reconstruct the required signal. alone and combined with a neural network. It has been con- 1718
1670 Samara et al. designed a virtual sensor to measure the AOA cluded from the analysis that the neural network models 1719
1671 based on Functional Pooling Non-linear AutoRegressive with increase the accuracy, but with the limitation of a complicated 1720
1672 eXogenous excitation (FP-NARX) methodology. It has been algorithm. The combined virtual AOA estimates the AOA 1721
1673 claimed that the FP-NARX methodology overcomes the limi- with an RMSE value of 0.12 degree which satisfies the design 1722
1674 tations of the conventional virtual sensor by capturing the requirement.83 1723
1675 non-linear aircraft dynamics and capturing the aircraft beha- Karl and Kamsan84 presented and compared the optimisa- 1724
1676 viour under different flight conditions. The MATLAB Simu- tion algorithms and neural networks as a virtual angle of 1725
1677 link was used to create the aircraft model. The MISO attack sensor. The Levenberg-Marquard (LM) algorithm was 1726
1678 NARX polynomial model structure is given by used to generate a map using the sensor data which can be 1727
1679 X X used to find the AOA. The LM is a standard algorithm used 1728
yðtÞ ¼ h0 þ hi1 ui1 ðtÞ þ hi1 i2 ui1 ðtÞui2 ðtÞ þ :::
to solve the non-linear least square problem. The algorithm 1729
i1 i1 ;i2
X optimises the parameters H as given in 1730
þ hi1 :::il ui1 ðtÞ:::uil ðtÞ þ eðtÞ ð32Þ 1731
1681 i1 ;:::;il
X
m
SðHÞ ¼ ½yi  fðxi ; HÞ2 ð35Þ
i¼1 1733
1682 Eq. (32) was modified, and FP-NARX model structure was
1683 developed, which can be used in all flight conditions. The FP- The LM algorithm used the aerodynamic model of the 1734
1684 NARX model was developed and tested for landing, takeoff UAV to simulate and estimate the AOA to fit the function 1735
and clean flight regimes. It has been concluded that the peak  
f V; a; b; g; Fpos ; H ! !
1685
a m (Fpos is the thrust lever position, g 1736
1686 error has not exceeded beyond 1.1° in all the flight regimes.82
1687 Marcel and Robert83 designed a virtual sensor for the mea- is the elevator deflection, H is the unknown parameter vector, 1737
!a m is the accelerometer sensor measurements). The UAVs 1738
1688 surement of AOA in small commercial aircraft. The combined
1689 white box linear time-varying model and grey box non-linear were used to get the in-flight data (T200 and ATTAS). The 1739

1690 model is used. Takagi-Sugeno (TS) fuzzy model and a black estimated values of AOA were compared with the actual 1740
AOA values. The second method was using the neural net- 1741

Table 3 AOA estimation results.


Da MSE Error variance
T200 simulation and LM optimization 0.0037 0.0036
T200 simulation and neural networks 0.0088 0.0088
ATTAS simulation and LM optimization 8.6  104 8.6  104
ATTAS simulation and neural networks 0.031 0.019
T200 flight data and LM optimization 1.011 0.1927
T200 flight data and neural networks 0.0390 0.1927

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
20 L. SANKARALINGAM, C. RAMPRASADH
1764
1742 works to estimate the AOA. The backpropagation method was kimax ¼ ki0 þ ki1 ðs=ðs þ 1ÞÞ i ¼ 1; 2; 3 ð36Þ 1766
1743 used for the learning by the network. This method adjusts the
The protected commanded signal nz,cA is formed by pro- 1767
1744 network weights as per the transfer function.
cessing the three load factor values. These protected values 1768
1745 From Table 3, it has been concluded that the developed vir-
allow the aircraft to fly within the safe flight envelop. The val- 1769
1746 tual sensor works with high accuracy and it can be imple-
idation of the protection law is shown in Fig. 24, in which all 1770
1747 mented in all the flight computers without any prior
the AOA sensors get faulty at t = 9.04 s. These results clearly 1771
1748 knowledge of the aircraft.84
show that the technique of alternate protection law produces 1772
1749 Hans-Dieter and Daniel85 developed a technique to
satisfactory results in the case of total AOA sensor failure.85 1773
1750 enhance flight control when there is a total AOA sensor loss.
1751 The author combined the modern Fault Detection and Diag-
1752 nosis (FDD) with a robust envelope protection algorithm. 6. Summary 1774

1753 The longitudinal protection law functionalities were used to


1754 avoid the state of practice degradation towards an unprotected This survey article intends to bring out clarity on all of the 1775
1755 control law. Fig. 23 shows the difficulty in controlling the air- available techniques for measurement of angle of attack 1776
1756 craft by the pilots when the AOA sensor fails. Table 4 sum- focused on mini flyers. The initial part of this article brings 1777
1757 marises the various types of faults that may occur in the out a flow diagram depicting various techniques available 1778
1758 AOA measurement and the strategy to overcome the faults. under the two major umbrellas. Measurement techniques 1779
1759 The FDD system monitors the sensors and isolates the involve mostly the employment of multi-hole probes which 1780
1760 faulty sensor. The value of a was taken from the last valid sig- boils down to pressure measurement. Angle of attack being a 1781
1761 nal in the case of total AOA sensor loss. The transfer matrices flow angle, which plays a very critical role in determining the 1782
1762 given in Eq. (36) were used to scale the difference in maximum aerodynamic forces of the aircraft, needs pressure measure- 1783
1763 and minimum ment at various angles and further calibration. These multi- 1784
hole probes based on pressure measurements require enough 1785
space to mount the probe as well as the instrumentation and 1786
air data computers. Further, it is an extremely complex task 1787
to calibrate such probes for low speed flights. Few such probes 1788
are commercially available, but at a huge cost due to the com- 1789
plexity involved in their calibration. Secondly, optical based 1790
sensors for flow angle measurement are also reported in the lit- 1791
erature. It has been observed that the optical sensors are used 1792
only for wind tunnel measurements. For in-flight flow angle 1793
measurements, especially for mini and micro flyers, optical 1794
methods prove unreliable. Meanwhile, potentiometer based 1795
Fig. 23 Illustration of FCS enhancement.85 vane probes are found suitable for applications in mini flyers. 1796
This article further throws light on the usage of estimation 1797
algorithms. But there is no well-developed estimation algo- 1798

Table 4 Fault accommodation strategy.


rithm available in the literature for applications in mini UAVs. 1799
Again, the estimation of flow angles proves to be another com- 1800
Scenario Running algorithm plex problem. In a nut shell, there is a need identified for fab- 1801
Fault free Nominal control law and protections rication and calibration of a suitable sensor for angle of attack 1802
1 or 2 AOA Fault isolation and use of remaining AOA measurement in mini UAVs. 1803
sensors fail sensor to maintain nominal control law and
protections Declaration of Competing Interest 1804
1 or 2 VCAS Fault isolation and use of remaining VCAS
sensors fail sensor to maintain nominal control law and
protections
The authors declare that they have no known competing 1805

3 AOA sensors Switch to alternate protections, maintaining financial interests or personal relationships that could have 1806

fail nominal longitudinal control law appeared to influence the work reported in this paper. 1807
3 VCAS sensors Switch to a VCAS independent longitudinal
fail control and stability augmentation system Acknowledgement 1808

We greatly acknowledge the financial support of the Aeronau- 1809


tical Research & Development Board (AR&DB) through the 1810
SIGMA Panel for sanctioning the project ID number 1811
ARDB/01/2021791/M/I. 1812

References 1813

1. Gracey W. Summary of methods of measuring angle of attack on 1814


aircraft. Washington, D.C.: NACA; 1958. Report No.: Technical 1815
Note 4351. 1816
Fig. 24 AOA response verification for full back stick command 2. Hermann PJ, Finley DB, Rehfeldt SC, Benishek LC. Airfoil probe 1817
input.85 for angle-of-attack measurement. J Aircraft 1984;21(1):87–9. 1818

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
Methods of angle of attack measurement and estimation in UAVs 21

1819 3. Kevin McDevitt T, Kevin Owen F. An optical angle of attack 24. Ramprasadh C, Arya H. Multistage-fusion algorithm for estima- 1887
1820 sensor. IEEE Aerosp Electron Syst Mag 1989;5(2):113–24. tion of aerodynamic angles in mini aerial vehicle. J Aircraft 1888
1821 4. Whitmore SA. Development of a pneumatic high-angle-of-attack 2011;49(1):93–100. 1889
1822 flush airdata sensing (HI-FADS) system. Washington, D.C.: 25. de La Parra S, Angel J. Low cost navigation system for UAV’s. 1890
1823 NACA; 1991. Report No.: NASA TM 104241. Aerosp Sci Technol 2005;9:504–16. 1891
1824 5. Wong DT. Evaluation of electrolytic tilt sensors for wind tunnel 26. Tapolcai D, Omran A, Newman B. Aircraft stall phenomenon 1892
1825 model angle-of-attack (AOA) measurements. International con- analysis using nonlinearity index theory. Aerosp Sci Technol 1893
1826 gress on instrumentation in aerospace simulation facilities. 1991. 2017;68:288–98. 1894
1827 6. Sergio C, Alessandro T, Michele Z, Alessandro G, Veronica R, 27. Wang XH, Cai LL. Aircraft navigation based on differentiation– 1895
1828 Marco T, et al. Aircraft angle of attack and air speed detection by integration observer. Aerosp Sci Technol 2017;68:109–22. 1896
1829 redundant strip pressure sensors. Proceedings of 2004 IEEE 28. Gäb A. Real-time parameter estimation for mini aerial vehicles 1897
1830 sensors. 2004 using low-cost hardware. Tag der mündlichen Prüfung 2012. 1898
1831 7. Park MH, Kim SS, Ryoo CK, Choi K, Park C. Development of 29. Lichota P, Lasek M. Maximum likelihood estimation: a method 1899
1832 alpha sensor for unmanned aerial systems. SICE annual confer- for flight dynamics – the angle of attack estimation. Proceedings of 1900
1833 ence 2008. 2008. the 14th International Carpathian Control Conference (ICCC). 1901
1834 8. Xu W, Wang M. The research of attack angle measurement of 2013. 1902
1835 projectile with CCD vertical target system. Proc Eng 30. Fekri MZ, Mobed M. Angle of attack estimation and aircraft 1903
1836 2011;15:4733–9. controller design using LQR methods. Canadian conference on 1904
1837 9. Mohammad S, Christian C. A hybrid system approach to electrical and computer engineering. 2005. 1905
1838 airspeed, angle of attack and sideslip estimation in unmanned 31. Johansen TA, Cristofaro A, Sørensen K, Hansen JM, Fossen TI. 1906
1839 aerial vehicles. 2015 International Conference on Unmanned On estimation of wind velocity, angle-of-attack and sideslip angle 1907
1840 Aircraft Systems (ICUAS). 2015. of small UAVs using standard sensors. 2015 International 1908
1841 10. Martos B, Rogers DF. Low-cost accurate angle-of-attack system. Conference on Unmanned Aircraft Systems (ICUAS). 2015. 1909
1842 J Aircraft 2018;55(2):660–5. 32. Telionis D, Yang YH, Rediniotis O. Recent developments in 1910
1843 11. Sumantra B, Akshay M. Angle of attack sensor using ball mouse multi-hole probe (MHP) technology. In 20th international con- 1911
1844 encoder wheel. 2016 International Conference on Intelligent gress of mechanical engineering. 2009. 1912
1845 Control Power and Instrumentation (ICICPI). 2016. 33. Nowack CFR. Improved calibration method for a five-hole 1913
1846 12. Wenz A, Johansen TA, Cristofaro A. Combining model-free and spherical pitot probe. J Phys E Sci Instrum 1970;3(1):1–26. 1914
1847 model-based angle of attack estimation for small fixed-wing UAVs 34. Wright MA. The evaluation of a simplified form of presentation 1915
1848 using a standard sensor suite. 2016 International Conference on for five-hole spherical and hemispherical pitometer calibration 1916
1849 Unmanned Aircraft Systems (ICUAS). 2016. data. J Phys E Sci Instrum 1970;3(5):356–62. 1917
1850 13. Vidya SB, Finitha KC, Jayanta D, Krishna S, Ramesh N, 35. Treaster AL, Yocum AM. The calibration and application of five- 1918
1851 Jayakumar M, et al. Differential pressure based angle of attack hole probes. Pennsylvania State University Park Applied Research 1919
1852 estimation in a Flush Air Data System (FADS). 2016 Interna- Lab; 1978. 1920
1853 tional Conference on Control, Instrumentation, Communication 36. Everett KN, Gerner AA, Durston DA. Theory and calibration of 1921
1854 and Computational Technologies (ICCICCT). 2016. non-nulling seven-hole cone probes for use in complex flow 1922
1855 14. Singh DJ, Dev R, Verma NK, Ghosh AK. Estimating angle of measurement. In AlAA 20th aerospace sciences meeting. Reston: 1923
1856 attack of an ATAS aircraft using TS fuzzy model. 2017 interna- AIAA; 1982. 1924
1857 tional conference on Sensing, Diagnostics, Prognostics, and 37. Ostowari C, Wentz Jr WH. Modified calibration technique of a 1925
1858 Control (SDPC). 2017. five-hole probe for high flow angles. Exp Fluids 1983;1(3):166. 1926
1859 15. Kushner LK, Drain BA, Schairer ET, Heineck JT, Bell JH. Model 38. Kjelgaard SO. Theoretical derivation and calibration technique of a 1927
1860 deformation and optical angle of attack measurement system in hemispherical-tipped, five-hole probe. Hampton: NASA Langley 1928
1861 the nasa unitary plan wind tunnel. 55th AIAA aerospace sciences Research Center; 1988. 1929
1862 meeting. Reston: AIAA; 2017. 39. Zilliac GG. Calibration of seven-hole pressure probes for use in fluid 1930
1863 16. Freeman DB. Angle of attack computation system. Wright- flows with large angularity. Moffett Field: Ames Research Center; 1931
1864 Patterson AFB, Ohio: Air Force Flight Dynamics Laboratory; 1989. 1932
1865 1973. Report No.: AFFDLTR-73-89. 40. Ligrani PM, Singer BA, Baun LR. Miniature five-hole pressure 1933
1866 17. Popowski S. Measurement and estimation of the angle of attack probe for measurement of three mean velocity components in low- 1934
1867 and the angle of sideslip. Aviation 2015;19(1):19–24. speed flows. J Phys E Sci Instrum 1989;22:868–76. 1935
1868 18. Olhausen J. Use of a navigation platform for performance 41. Depolt T, Koschel W. Investigation on optimizing the design 1936
1869 instrumentation on the YF-16. AIAA 13th aerospace sciences process of multi-hole pressure probes for transonic flow with panel 1937
1870 meeting. Reston: AIAA; 1975. methods. 1991 international congress on instrumentation in 1938
1871 19. Thacker T. Use of state estimation to calculate angle of attack aerospace simulation facilities. 1991. 1939
1872 position error from flight test data[dissertation]. Wright-Patterson 42. Dominy RG, Hodson HP. An investigation of factors influencing 1940
1873 AFB, Ohio: Air Force Institute of Technology; 1985. the calibration of five-hole probes for three-dimensional flow 1941
1874 20. Zeis Jr JE. Angle of attack and sideslip estimation using an inertial measurements. J Turbomach 1993;115(3):513–9. 1942
1875 reference platform [dissertation]. Wright-Patterson AFB, 43. Hooper JD, Musgrove AR. Reynolds stress, mean velocity, and 1943
1876 Ohio: Air Force Institute of Technology; 1988. dynamic static pressure measurement by a four-hole pressure 1944
1877 21. Ramprasadh C, Arya H. Multistage-fusion algorithm for estima- probe. Exp Therm Fluid Sci 1997;15(4):375–83. 1945
1878 tion of aerodynamic angles in mini aerial vehicle. J Aircraft 44. Fingersh L, Robinson M, Fingersh L, Robinson M. Wind 1946
1879 2012;49(1):93–100. tunnel calibration of 5-hole pressure probes for application to 1947
1880 22. Long H, Song SJ. Method of estimating angle-of-attack and wind turbines. In 35th aerospace sciences meeting and 1948
1881 sideslip angel based on data fusion. 2009 second international exhibit.1997. 1949
1882 conference on intelligent computation technology and automa- 45. Morrison GL, Schobeiri MT, Pappu KR. Five-hole pressure 1950
1883 tion. 2009. probe analysis technique. Flow Meas Instrum 1998;9:153–8. 1951
1884 23. Ramprasadh C, Arya H. Estimation of aerodynamic angles in a 46. Li YH. Numerical investigation of the influence of Reynolds 1952
1885 mini aerial vehicle under turbulent atmosphere. AIAA atmo- number on probe measurements. Tsinghua Sci Technol 2000. 1953
1886 spheric flight mechanics conference. Reston: AIAA; 2011.

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003
CJA 1435 No. of Pages 22
21 November 2019
22 L. SANKARALINGAM, C. RAMPRASADH

1954 47. Babinsky H, Kuschel U, Hodson HP, Moore DF, Welland ME. 66. Milanovic IM, Kalkhoran IM. Numerical calibration of a conical 2013
1955 Geometries for five-hole-type probes with planar sensor arrays. five-hole probe for supersonic measurements. Meas Sci Technol 2014
1956 AIAA J 2001;39(12):2414–6. 2000;11(12):1812–8. 2015
1957 48. Robert Porro A. Pressure probe designs for dynamic pressure 67. Wildmann N, Ravi S, Bange J. Towards higher accuracy and 2016
1958 measurements in a supersonic flow field. 19th International better frequency response with standard multi-hole probes in 2017
1959 congress on instrumentation in aerospace simulation facilities; turbulence measurement with remotely piloted aircraft (RPA). 2018
1960 2001. Atmos Meas Tech 2014;7:1027–41. 2019
1961 49. Pisasale AJ, Ahmed NA. A novel method for extending the 68. Pisasale AJ, Ahmed NA. Theoretical calibration for highly three- 2020
1962 calibration range of five-hole probe for highly three-dimensional dimensional low-speed flows of a five-hole probe. Meas Sci 2021
1963 flows. Flow Meas Instrum 2002;13:23–30. Technol 2002;13:1100–7. 2022
1964 50. Parameswaran V, Jategaonkar R. Five-hole flow probe calibration 69. Jadhav VB, Sarode PL, Salunke NP, Baviskar PV. Fabrication 2023
1965 from dynamic and tower flyby maneuvers. J Aircraf 2005;42 and calibration of five hole pressure probe. Int J Innovat Res Sci 2024
1966 (1):80–6. Eng Technol 2016;5(4):4880–7. 2025
1967 51. Lee SW, Sang BJ. Reynolds number effects on the non-nulling 70. Bereznai J, Mlynár P. The calibration of five-hole pressure probe. 2026
1968 calibration of a cone-type five-hole probe for turbomachinery In AIP conference proceedings 1768. 2016. 2027
1969 applications. J Mech Sci Technol 2005;19(8):1632–48. 71. Ratneswar M, Chowdhury D, Sinha PK, Majumdar B. Calibra- 2028
1970 52. Zhai J, Rebstock R, Hufnagel K. New method for evaluating the tion of a five-hole probe in null and non-null technique. Int J Eng 2029
1971 hemispheric five-hole probes. J Aircraft 2006;43(3):855–8. Res Appl 2016;6(5):27–32. 2030
1972 53. Caldeira-Pires A, Neiva RQ, de Mello Pereira R. Experimental 72. Shevchenko AM, Shmakov AS. Multi-hole pressure probes to 2031
1973 study of a five-hole pressure probe for measuring three mean wind tunnel experiments and air data systems. In AIP conference 2032
1974 velocity components. In 19th international congress of mechanical proceedings 1893. 2017. 2033
1975 engineering. 2007. 73. Conlon MJ, Wright A, Abo El Ella HM. Measurement of large 2034
1976 54. Alberto C, Roberto G, Veronica P, Francesco S. Multi-hole probe flow angles with non-nulling multi-hole pressure probes. In 2035
1977 and elaboration algorithms for the reconstruction of the air data Proceedings of ASME Turbo Expo 2017: turbomachinery techni- 2036
1978 parameters. 2008 IEEE international symposium on industrial cal conference and exposition. 2017. 2037
1979 electronics. 2008. 74. Benjamin H, Thomas P. The Oxford probe: an open access five- 2038
1980 55. Argüelles Dı́az KM, Fernández Oro JM, Blanco Marigorta E. hole probe for aerodynamic measurements. Meas Sci Technol 2039
1981 Cylindrical three-hole pressure probe calibration for large angular 2017;28 035004. 2040
1982 range. Flow Meas Instrum 2009;20:57–68. 75. Daniel W, Tadeusz C, Janusz K. Experimental and numerical 2041
1983 56. Kim SH, Kang YJ, Myong RS, Cho TH, Park YM, Choi IH. investigations of the averaging Pitot tube and analysis of instal- 2042
1984 Calibration of a five-hole multi-function probe for helicopter air lation effects on the flow coefficient. Flow Meas Instrum 2008;19 2043
1985 data sensors. Int J Aeronaut Space Sci 2009;10(2):43–51. (5):301–6. 2044
1986 57. Pavel P, Karel D, Tomáš Č, Vı́tězslav H, Ondřej V. A combined 76. Benjamin OJ. A multi-holed pressure probe accuracy analysis 2045
1987 angle of attack and angle of sideslip smart probe with twin [dissertation]. Florida: University of Florida; 2012. 2046
1988 differential sensor modules and doubled output signal. In 77. Tolga Y, Guillermo P. Robust procedure for multi-hole probe 2047
1989 Proceedings of 2010 IEEE sensors. 2010. data processing. Flow Meas Instrum 2012;26:46–54. 2048
1990 58. Argüelles Dı́az KM, Fernández Oro JM, Blanco Marigorta E, 78. Paul C, Ethan D, Daniel F, Stephen K. Design and construction of 2049
1991 Barrio Perotti R. Head geometry effects on pneumatic three-hole a velocity probe calibration rig. Worcester: Worcester Polytechnic 2050
1992 pressure probes for wide angular range. Flow Meas Instrum Institut; 2013. 2051
1993 2010;21:330–9. 79. Nekkanti S, Kancherla S. Effect of chamfer angle on the 2052
1994 59. Akshoy RP, Ravi RU, Anuj J. A novel calibration algorithm for calibration curves of five hole probes. Int J Rotating Mach 2053
1995 five-hole pressure probe. Inte J Eng Sci Technol 2011;3(2):89–95. 2014;704315:1–11. 2054
1996 60. Lien SJ, Ahmed NA. An examination of suitability of multi-hole 80. Bartosz B, Zdzisław R, Konrad W, Bartosz G, Krzysztof K. 2055
1997 pressure probe technique for skin friction measurement in turbu- Miniature airflow probe for an unmanned aerial vehicle. 2016 2056
1998 lent flow. Flow Measur Instrument 2011;22(3):153–64. IEEE metrology for aerospace (MetroAeroSpace). 2016. 2057
1999 61. Campbell JF, Brandon JM. Calibration and flight results for the 81. Fatma CS, Schiffmann J. Dynamic pressure probe response tests 2058
2000 Ares I-X 5-hole probe. Acta Astronaut 2011;68:1219–27. for robust measurements in periodic flows close to probe resonat- 2059
2001 62. Crowley C, Shinder II, Moldover MR. The effect of turbulence on ing frequency. Meas Sci Technol 2018;29 025301. 2060
2002 a multi-hole pitot calibration. Flow Meas Instrum 2013;33:106–9. 82. Samara PA, Fouskitakis GN, Sakellariou JS, Fassois SD. Aircraft 2061
2003 63. Ravirai J, Nekkanti S, Gajanan C. A miniature four-hole probe angle-of-attack virtual sensor design via a functional pooling narx 2062
2004 for measurement of three-dimensional flow with large gradients. methodology. 2003 European control conference (ECC). 2003. 2063
2005 Int J Rotat Mach 2014;2014:1–12, 297861. 83. Marcel O, Robert B. Virtual sensor for the angle-of-attack signal 2064
2006 64. Georgiou DP, Milidonis KF. Fabrication and calibration of a sub- in small commercial aircraft. IEEE international conference on 2065
2007 miniature 5-hole probe with embedded pressure sensors for use in fuzzy systems. 2006. 2066
2008 extremely confined and complex flow areas in turbomachinery 84. Karl K, Kamsan S. Application and comparison of neural 2067
2009 research facilities. Flow Meas Instrum 2014;39:54–63. networks and optimization algorithms as a virtual angle of attack 2068
2010 65. Lee HH, Shinder II, Wright JD, Moldover MR. Application of sensor. 2013 IEEE international conference on computational 2069
2011 ANFIS method to the non-nulling calibration of multi-hole pitot intelligence and cybernetics (CYBERNETICSCOM). 2013. 2070
2012 tube. Procedia Eng 2014;79:125–32. 85. Hans-Dieter J, Daniel O. Enhancing flight control in case of total 2071
angle of attack sensor loss. 2016 3rd conference on control and 2072
fault-tolerant systems (SysTol). Barcelona: Spain; 2016. 2073
2074

Please cite this article in press as: SANKARALINGAM L, RAMPRASADH C A comprehensive survey on the methods of angle of attack measurement and esti-
mation in UAVs, Chin J Aeronaut (2019), https://doi.org/10.1016/j.cja.2019.11.003

You might also like