29494 IEEE SENSORS JOURNAL, VOL. 23, NO.

23, 1 DECEMBER 2023

Ultrathin Flexible Skin With All-Polyimide

Pressure and Airflow Sensor Array for
Estimation of Flight Parameters
Zihao Dong , Zheng Gong , Bangqi Chen, Tianyu Sheng, Yudong Cao, Qipei He,
Peng Zhao, and Yonggang Jiang

Abstract—The estimation of the airspeed and angle of

attack (AOA) is crucial for achieving precise flight con-
trol in small unmanned aerial vehicles. A flexible skin with
distributed flow sensors has been proven to be an effec-
tive approach to estimate flight parameters, particularly,
pressure-airflow data fusion strategies can improve estima-
tion accuracy. For the first time, we propose an ultrathin
(70 µm) all-polyimide flexible skin integrated with capacitive
pressure sensors and piezoresistive airflow sensors. The
ultrathin structure of the flexible skin endows it with flexibility
and adaptability to curved airfoils and avoids disturbing the flow field near the surface. Meanwhile, the rational
arrangement of the sensors on the leading edge of the airfoil makes them insensitive to the angle of sideslip (AOS).
The pressure and airflow sensors achieve a resolution of 4.5 Pa and 2.4 m/s, respectively. Using a simple multilayer
perceptron (MLP) neural network, we estimate the average solving errors of airspeed and AOA to be 0.79 m/s and
0.67◦ , respectively. Compared with traditional methods, the proposed flexible skin considerably reduces the workload by
decoupling the airspeed, AOA, and AOS, providing a facile method for efficient and accurate flight parameter estimation.
Index Terms— Airflow sensor, flexible electronics, piezoresistive cantilever, pressure sensor.

I. I NTRODUCTION incorporate into SUAVs due to their large size and poor surface
MALL unmanned aerial vehicles (SUAVs) have emerged adaptability [11], [12], [13], [14], [15]. With the advent of
S as valuable tools for various applications, including dis-
aster relief, meteorology, and logistics transportation [1], [2],
micro-electromechanical system (MEMS) technology, flexible
skins based on MEMS sensor arrays have become important
[3]. However, their small size renders them vulnerable to gusty tools for estimating the flight parameters of SUAVs.
winds and turbulence, making flight control a major challenge With the aim of high-reliability and low-cost flight control
[4], [5]. Estimating flight parameters such as airspeed, angle of SUAVs, various flexible skins integrated with pressure or
of attack (AOA), and angle of sideslip (AOS) is essential airflow sensors were proposed. Pang et al. [16] suggested
for the control of aircraft [6], [7], [8], [9], [10]. Traditional a flexible skin that employed thermal flow and commercial
AOA sensors and pitot tubes have proven challenging to pressure sensors to detect the boundary layer separation.
Meanwhile, a flexible skin that encompassed commercial
Manuscript received 26 September 2023; accepted 12 October
2023. Date of publication 20 October 2023; date of current version
pressure and flow sensors was reported, targeting airspeed and
30 November 2023. This work was supported by the National Natural AOA. However, it did not consider the effects of AOS [17].
Science Foundation of China under Grant 52022008, Grant 51975030, These studies relied on commercial rigid sensors that were
and Grant T2121003. The associate editor coordinating the review of
this article and approving it for publication was Dr. Yang Yang. (Corre-
challenging to fit appropriately to the airfoil surface and are
sponding author: Yonggang Jiang.) typically thick (∼1 mm), which might disturb the flow field
Zihao Dong, Zheng Gong, Bangqi Chen, Tianyu Sheng, Yudong Cao, near the surface. To address this concern, Xiong et al. [18]
Qipei He, and Yonggang Jiang are with the Institute of Bionic
and Micro-Nano Systems, School of Mechanical Engineering and
developed a polyimide (PI) flexible skin for multifunctional
Automation, Beihang University, Beijing 100191, China (e-mail: flying perception, including surface pressure, temperature, and
jiangyg@buaa.edu.cn). wall shear stress. However, they failed to provide solutions for
Peng Zhao is with the Institute of Bionic and Micro-Nano Systems,
School of Mechanical Engineering and Automation, Beihang University,
AOA and airspeed. Therefore, flexible sensing skin technology
Beijing 100191, China, and also with the Beijing Institute of Space must be further investigated to estimate the flight parameters
Launch Technology, Beijing 100076, China. of SUAVs.
This article has supplementary downloadable material available at
https://doi.org/10.1109/JSEN.2023.3324809, provided by the authors.
In this study, an ultrathin flexible skin based on pressure
Digital Object Identifier 10.1109/JSEN.2023.3324809 and airflow sensors was introduced to derive airspeed and

1558-1748 © 2023 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: NANJING UNIVERSITY OF AERONAUTICS AND ASTRONAUTICS. Downloaded on May 13,2025 at 13:53:01 UTC from IEEE Xplore. Restrictions apply.
 DONG et al.: ULTRATHIN FLEXIBLE SKIN WITH ALL-POLYIMIDE PRESSURE AND AIRFLOW SENSOR ARRAY 29495

Fig. 1. Schematic of flexible skin and sensing principle: (a) SUAV perceive flight parameters using flexible skin; (b) three-layer polyimide structure
of the sensor array; and (c) sensing principle of airflow sensor and pressure sensor.

AOA using a neural network model. This was achieved by the pressure sensors, the electrodes of the airflow sensors,
the distributed sensing of the pressure and airflow along the piezoresistors, and the cantilever beams. The cavities and
the leading edge of the SUAV airfoil. The skin comprises connecting microchannel of the pressure sensors and partial
capacitive pressure sensors, piezoresistive airflow sensors, cavities of the airflow sensors were designed in the middle
and the corresponding temperature compensation units. The layer. Similarly, the bottom electrodes of the pressure sensors
all-polyimide construction of the ultrathin, flexible skin pro- and the other parts of the cavities of the airflow sensors were
vides considerable flexibility and is perfectly adapted to the designed to be in the bottom layer. Five capacitive pressure
curved shape of the airfoil without disturbing the flow field. sensors and ten piezoresistive velocity sensors comprised the
Wind tunnel experiments demonstrated that the calculation flexible skin. The sensing principle of pressure and airflow
errors for the AOA and airspeed were 0.67◦ and 0.79 m/s, sensors is schematically illustrated in Fig. 1(c).
respectively. Moreover, the calculations were not influenced The relevant calculation for the airflow sensor was reported
by the change of AOS, significantly reducing the tasks of in [19] and the average strain of the piezoresistor is expressed
decoupling AOA, airspeed, and AOS. as
LR
9 ρV 2 L 3
Z
II. D ESIGN AND C ALCULATION 1 9q L
ε= (L − x)dx = (1)
Fig. 1(a) depicts the application and structure of the flexible LR 0 4bh 2V E 16 Eh 2V L R
skin in a schematic. The flexible skin was mounted on the
leading edge of the airfoil to maximize the accuracy of where ε is the average strain, and L, h V , b, and E are
the flight parameter evaluation [9]. The pressure and flow the length, thickness, width, and Young’s modulus of the
velocity distributions at the leading edge of the wing were cantilever, respectively. L R is the linear length of piezoresistor
detected by the flexible skin to calculate the airspeed and AOA. distribution on the cantilever beam, ρ is the air density, and
The pressure sensors were connected through a microchannel q is the uniform load on the cantilever. Fig. S1(a) illustrates
to ensure that the internal reference pressures of the five the variation in the average strain of the piezoresistor as a
sensors remained consistent. The airflow sensor featured a flat function of the airflow sensor dimensions at different flow
cantilever beam integrated with a piezoresistor that responded velocities. The average strain increases quadratically as the
to both the velocity and direction of the incoming airflow. length of the cantilever beam increases. However, the average
Moreover, temperature compensation elements were employed strain decreases as the piezoresistor length and cantilever
to eliminate the impact of the ambient temperature. Fig. 1(b) thickness increase. Considering the fabrication process and
shows that the flexible skin comprises a trilayer polyimide device sensitivity, the sensor size was determined, as listed
structure. The upper layer contained the upper electrodes of in Table I.

Authorized licensed use limited to: NANJING UNIVERSITY OF AERONAUTICS AND ASTRONAUTICS. Downloaded on May 13,2025 at 13:53:01 UTC from IEEE Xplore. Restrictions apply.
 29496 IEEE SENSORS JOURNAL, VOL. 23, NO. 23, 1 DECEMBER 2023

TABLE I
D IMENSIONS OF THE D ESIGNED A IRFLOW AND P RESSURE S ENSOR

Fig. 2. Fabrication sequence of the flexible skin. (a) PI spin coating. (b) Pizeoresistors depositing. (c) Electrodes depositing. (d) Al depositing.
(e) PI spin coating. (f) Al depositing. (g) PI etching. (h) Al etching. (i) PI spin coating. (j) Electrodes depositing. (k) PI spin coating and etching.
(l) Laser etching. (m) PI bonding. (n) Si removing. (o) PI etching.

The pressure sensor was designed with a parallel-plate Cgap can be rewritten as
configuration [20]. The capacitor between the upper and
44.26a 2 d
bottom electrodes can be regarded as two capacitors in series: Cgap = ln . (6)
capacitor CPI with PI as the medium and capacitor Cgap with Y d − 0.628Y
air as the medium. The following relationship can be obtained Here, Y is given by Y = (1 Pa 4 /Eh P )1/3 , where 1P is
from the formula for capacitors in series: the differential pressure and h P is the thickness of the PI
membrane.
1 1 1
= + (2) By combining (2)–(6), the relationship between capacitance
C CPI Cgap and each pressure chamber size can be derived, as plotted
where CPI and Cgap can be calculated as in Fig. S1(b). The variation in capacitance increases with an
increase in the film radius and decreases with an increase in
πε0 εPI a 2 the depth of the pressure cavity and the thickness of the PI
CPI = (3)
d membrane. For optimal performance in measuring pressures
ε0r
Z 2π Z a
within the range of 1000 Pa, the final sensor dimensions were
Cgap = dr dθ. (4)
0 0 d − ω(r ) determined, as listed in Table I.
Here, ε0 and εPI are the permittivity of vacuum and relative
III. FABRICATION P ROCESS
dielectric constant of the PI film, respectively, r is the distance
from the center of the chamber, and a and d are the radius A. Fabrication of the Flexible Skin
and depth of the pressure chamber, respectively. ω(r ) is the Flexible skin was fabricated using MEMS technology [21].
local deflection, which can be defined as The fabrication process of the flexible skin is shown in Fig. 2.
1  A silicon wafer was covered with a layer of 5-µm-thick PI film
1Pa 4 3 r2
 
[Fig. 2(a)]. The constantan piezoresistor was prepared using a
ω(r ) ≈ 0.628 1− 2 . (5)
Eh a lift-off process [Fig. 2(b)]. Subsequently, the chromium-gold

Authorized licensed use limited to: NANJING UNIVERSITY OF AERONAUTICS AND ASTRONAUTICS. Downloaded on May 13,2025 at 13:53:01 UTC from IEEE Xplore. Restrictions apply.
 DONG et al.: ULTRATHIN FLEXIBLE SKIN WITH ALL-POLYIMIDE PRESSURE AND AIRFLOW SENSOR ARRAY 29497

Fig. 3. Final device: (a) photograph of the flexible skin with pressure and airflow sensor arrays, and the inset is the enlarged view of the airflow
sensor and (b) flexible skin attached to an airfoil model.

electrode was fabricated using the same method [Fig. 2(c)]. During the calibration of the pressure sensors, the working
An aluminum sacrificial layer (100 nm) was sputtered to pressure of the wind tunnel was introduced into the microchan-
serve as a mask for the subsequent etching process of the nel connecting the pressure sensors via a guide pipe, acting
cantilever beam and as a protective layer for the pressure as the reference pressure [22], and the atmospheric pressure
chamber [Fig. 2(d)]. Then, a 30-µm-thick layer of PI was acted as the applied pressure to the pressure sensors. The
spin-coated as a middle layer [Fig. 2(e)]. A layer of 200-nm differential pressure was tuned by changing the wind tunnel
aluminum was sputtered, and the photoresist was used as the velocity and quantified using a commercial pressure sensor.
mask. Wet etching with a KOH solution (0.5% wt.) was used An inductance–capacitance–resistance (LCR) meter was used
to form a patterned aluminum mask [Fig. 2(f)]. Subsequently, to simultaneously collect capacitance signals from the pressure
oxygen plasma was used to etch the cantilever beam, pressure sensors. The airflow sensors situated on the flat base were
chamber, and microchannel [Fig. 2(g)]. Finally, the remain- directly tested in a wind tunnel, and a digital source meter
ing Al sacrificial layer was removed using a KOH solution was used to collect their voltage outputs. Fig. 4(a) and (b)
[Fig. 2(h)]. shows the dynamic response of pressure and airflow sensors
To process the bottom structure, a 30-µm-thick PI was under a reciprocating pressure of 370 Pa and reciprocating
spin-coated on the silicon wafer [Fig. 2(i)], followed by the airflow velocity of 25 m/s, exhibiting good repeatability. The
fabrication of the electrodes using a lift-off process [Fig. 2(j)]. capacitance output of the pressure sensor increases as the
A layer of 5-µm-thick PI was spin-coated as an insulating pressure increases, as shown in Fig. 4(c). By linearly fitting
layer to prevent the upper and bottom electrodes of the pres- the sensor output and pressure and considering the noise level
sure sensors from contacting short circuit [Fig. 2(k)]. Finally, of the sensor, the resolutions of the four pressure sensors were
the cavity of the airflow sensor was etched using a CO2 laser calculated as 11.3, 11.1, 9.2, and 4.5 Pa, respectively. Simi-
[Fig. 2(l)]. larly, the resolution of the five airflow sensors was calculated
The upper and bottom layers were assembled using an as 3.1, 3.2, 2.8, 2.4, and 2.4 m/s, respectively, by establishing
epoxy resin as an adhesive and cured at room temperature a linear fit between the output and squared velocity of the
for 24 h [Fig. 2(m)]. Then, the device was stripped from the airflow sensors, as shown in Fig. 4(d).
silicon substrate [Fig. 2(n)] and the plasma was etched to As it takes a long time for the wind tunnel to reach a stable
expose the pins and complete the processing of the flexible state, directly determining the response time of the sensor in a
skin [Fig. 2(o)]. wind tunnel is difficult. A solenoid valve was used to control
the high-pressure gas injection directly into the sensors to
B. Final Device measure the response time of the sensor accurately. As shown
The flexible sensing skin was successfully fabricated, in Fig. 4(e) and (f), the response times of the pressure and
as shown in Fig. 3(a), comprising five pressure sensors, ten airflow sensor are 0.39 and 0.26 s, respectively. A long-term
airflow sensors, and two temperature compensation units. The cyclic experiment was performed, and the sensors P5 and V5
enlarged view of the airflow sensor shows the cantilever beam showed good stability after 300 cycles (Fig. S3).
with the piezoresistor for flow velocity sensing. The flexible
skin was assembled on the leading edge of the 3-D-printed
airfoil with a curvature radius of 20 mm, as shown in Fig. 3(b). B. Characterization of Flexible Skin in Wind Tunnel
The experimental setup, as shown in Fig. 5(a), was con-
IV. R ESULTS AND D ISCUSSION structed to evaluate the capability of the flexible skin to
A. Sensor Characterization estimate the flight parameters. A 3-D-printed airfoil, integrated
We calibrated the flexible skin before integrating it into the with the flexible skin, was inserted into the wind tunnel, and
airfoil because the angle between the cantilever-based airflow connected to a rotating micropositioner. This manipulation
sensor and the incoming flow is affected by the curved leading enables precise control of the AOA, with variation ranging
edge. A calibration setup was built to assess the performance between −20◦ and 20◦ . Additionally, a microrotator was
of the pressure and airflow sensors, as shown in Fig. S2(a). installed on the airfoil to manually obtain the AOS conditions

Authorized licensed use limited to: NANJING UNIVERSITY OF AERONAUTICS AND ASTRONAUTICS. Downloaded on May 13,2025 at 13:53:01 UTC from IEEE Xplore. Restrictions apply.
 29498 IEEE SENSORS JOURNAL, VOL. 23, NO. 23, 1 DECEMBER 2023

Fig. 4. Calibration results of pressure and airflow sensors: (a) output of pressure sensors under periodic pressure of 370 Pa; (b) output of airflow
sensors under periodic flow velocity of 25 m/s; (c) output of pressure sensor variation with pressure; (d) output of airflow sensor variation with flow
velocity; (e) response time of the pressure sensor P5; and (f) response time of the airflow sensor V5.

of 10◦ and 20◦ , as depicted in Fig. 5(b). The signals obtained component. At 10◦ and 20◦ AOSs, the normal incoming
from the sensors are used to train a neural network that component became 98% and 94% of the valve of 0◦ AOS,
ultimately estimates the flight parameters, such as the AOA respectively. In comparison, the incoming shear component
and airspeed at different AOSs. was 17% and 34% of the original flow field, respectively.
The wind tunnel experiment shows the variation of the During the experiment, when the incoming flow velocity
output of the eight sensors under different airspeeds and AOSs was 35 m/s and the AOS was 20◦ , the normal component
with changes in the AOA, as illustrated in Fig. 5(c) and (d). of the incoming flow was 33 m/s. Therefore, the output
The local pressure readings of the pressure sensors were signals of the airflow sensors exhibited negligible variation
closely linked to the sensor locations at the leading edge, under different AOSs. Additionally, the airflow sensors did not
as illustrated in Fig. S4. During the gradual increase of AOA response to the shear component of the incoming flow. With
from −20◦ to 20◦ , the stagnation point on the leading edge respect to the pressure sensors, the reference pressure was the
progressively shifts closer to P1 and P2 while moving away internal pressure of the airfoil because the microchannel was
from P4 and P5. Consequently, the local pressure of P1 and connected to the inside of the airfoil. The shear component
P2 increased, while the pressure of P4 and P5 decreased of the incoming flow did not change the internal reference
throughout the process. This leads to a gradual increase in the pressure. Moreover, the wind-induced pressure on the surface
capacitances of sensors P1 and P2, whereas the capacitances of the sensing points changed negligibly, as determined by the
of sensors P4 and P5 gradually decrease. In the case of airflow slightly varying normal component. Therefore, the variation in
sensors, as the AOA of the wing slowly increases from −20◦ the AOS had a marginal influence on the output of the airflow
to 20◦ , the cantilever beam’s angle to the incoming airflow and pressure sensors.
increases, leading to an increase in cantilever deflection and
resistance of the airflow sensors V1 and V2. In contrast, the C. Estimation of Flight Parameters
angle between the cantilever beam of sensor V4 (or V5)
A multilayer perceptron (MLP) neural network is a feed-
and the incoming airflow gradually decreased, leading to a
forward neural network model that comprises an input layer,
decrease in the cantilever deflection and a decline in the output
multiple hidden layers, and an output layer [23], as shown in
of the sensor.
Fig. 6(a). Due to its adaptive learning ability, robustness to
Notably, the output signals of the eight sensors changed
noise, and simple structure [24], the MLP neural network was
only marginally for different AOSs. The AOS change reduced
utilized to train the test data and estimate the flight parameters
the normal incoming component, resulting in a larger shear
in this study. The pressure and airflow data serve as the input

Authorized licensed use limited to: NANJING UNIVERSITY OF AERONAUTICS AND ASTRONAUTICS. Downloaded on May 13,2025 at 13:53:01 UTC from IEEE Xplore. Restrictions apply.
 DONG et al.: ULTRATHIN FLEXIBLE SKIN WITH ALL-POLYIMIDE PRESSURE AND AIRFLOW SENSOR ARRAY 29499

Fig. 5. Wind tunnel experimental results of the pressure/airflow sensor array: (a) diagram of experimental setup; (b) angle variation of the airfoil, the
AOA changed from −20◦ to 20◦ (with step of 4◦ ), and the AOS was set to 0◦ , 10◦ , and 20◦ , respectively; (c) capacitance variation of four pressure
sensors under different AOAs, AOSs, and airflow velocities; and (d) outputs of four flow sensors under different AOAs, AOSs, and airflow velocities.

TABLE II
S TUDIES ON F LEXIBLE S KIN

layer, and the MLP neural network outputs the airspeed and represents the solution using data from only the four pressure
AOA after the three-layer neural network training and fitting. sensors, the red circle represents the solution using data from
All the test data of different AOSs (0◦ , 10◦ , and 20◦ ) were the four airflow sensors, and the blue triangle represents the
trained together. Data from the same AOS were divided into solution based on data from both the four pressure sensors
55 data points based on the airspeed and AOA, and we selected and the four airflow sensors. Additionally, the errors of the
ten random points from each AOS, for a total of 30, to form ten points were averaged to arrive at the average errors of
the test set; the remaining 135 points were used as the training the airspeed and AOA assessed, as shown in Fig. 6(e). The
set, as illustrated in Fig. 6(b). This process helped validate experimental results reveal that the average solving error of
the effectiveness of the MLP neural network in estimating the the pressure-airflow fusion method is substantially smaller than
airspeed and AOA while neglecting the AOS. those of the other two methods, which is consistent with [17].
The calculated errors were obtained by comparing the For different AOSs, the average airspeed solving errors are
results with the actual values of the test set to determine 0.73, 0.79, and 0.78 m/s, while the average AOA calculation
the estimation accuracy of both the airspeed and the AOA. errors are 0.66◦ , 0.63◦ , and 0.67◦ , respectively. The evaluation
Fig. 6(c) and (d) shows the airspeed and AOA solving errors of results confirm the capability of flexible skin to evaluate flight
the test points at different AOSs. Specifically, the black square parameters with acceptable accuracy. Moreover, the consistent

Authorized licensed use limited to: NANJING UNIVERSITY OF AERONAUTICS AND ASTRONAUTICS. Downloaded on May 13,2025 at 13:53:01 UTC from IEEE Xplore. Restrictions apply.
 29500 IEEE SENSORS JOURNAL, VOL. 23, NO. 23, 1 DECEMBER 2023

Fig. 6. Flight parameter calculation results using artificial neural network: (a) diagram of MLP neural network and (b) schematic of training set and
test set. (c) Absolute velocity error and (d) absolute AOA error of ten test points estimated with pure pressure data (black square), pure velocity
data (red circle), and pressure/velocity fusion data (blue triangle) at different AOSs. (e) Mean absolute error of estimated airspeed and AOA using
different data at different AOSs.

error at different AOSs confirms that the AOS has a negligible proposed method, the flexible skin is attached to the leading
effect on the estimation of the flight parameters using the edge rather than to the upper surface of the airfoil, making it
flexible skin. sensitive to airspeed and AOA.
Compared with the related studies, the flexible skin in this
V. C ONCLUSION study possesses an ultrathin profile (70 µm) to achieve high
We proposed a flexible skin for measuring the airspeed flexibility due to the all-polyimide structure. The flexibility of
and AOA of SUAVs. The ultrathin flexible skin was designed the skin enables it to adapt to various airfoil surfaces. Addi-
to incorporate capacitive differential pressure sensors and tionally, the reduced thickness avoids disturbing the flow field
piezoresistive airflow sensors, which were batch-fabricated near the surface of the airfoil, which is a critical aspect of the
using MEMS processes. The pressure and airflow sensors flight dynamics of SUAVs. Moreover, the estimation accuracy
exhibited a minimum resolution of 4.5 Pa and 2.4 m/s, of the airspeed and AOA is acceptable. In conclusion, the
respectively. Subsequently, the flexible skin was integrated into flexible skin proposed in this study offers unique advantages
the leading edge of the airfoil to measure the pressure and over related studies for precise AOA and airspeed measure-
airflow distribution, and estimate the flight parameters via an ment of SUAVs. Future improvements in sensor performance
MLP neural network. The experimental results demonstrated can enhance the solving accuracy, paving the way for flexible
that the solving errors for the AOA and airspeed were 0.67◦ skin-based flight parameter estimation.
and 0.73 m/s, respectively.
Table II lists the relevant studies on the flexible skin of R EFERENCES
SUAV. From a structural perspective, most flexible skins inte- [1] H. Shen, N. Li, H. Griffiths, and S. Rojas, “Tracking control
of a small unmanned air vehicle with airflow awareness,” in
grate rigid sensors onto a flexible substrate. These rigid sensors Proc. Amer. Control Conf. (ACC), May 2017, pp. 4153–4158, doi:
are generally thick, which may disturb the flow field near the 10.23919/ACC.2017.7963593.
surface and reduce the flexibility of the substrate. In terms [2] A. Mark, Y. Xu, and B. T. Dickinson, “Review of microscale
flow-sensor-enabled mechanosensing in small unmanned aerial vehi-
of function, not all flexible skins can estimate the airspeed cles,” J. Aircr., vol. 56, no. 3, pp. 962–973, May 2019, doi:
and AOA, which are related to the airfoil arrangement. In the 10.2514/1.C034979.

Authorized licensed use limited to: NANJING UNIVERSITY OF AERONAUTICS AND ASTRONAUTICS. Downloaded on May 13,2025 at 13:53:01 UTC from IEEE Xplore. Restrictions apply.
 DONG et al.: ULTRATHIN FLEXIBLE SKIN WITH ALL-POLYIMIDE PRESSURE AND AIRFLOW SENSOR ARRAY 29501

[3] H. Shen, Y. Xu, and B. T. Dickinson, “Fault tolerant attitude control for [24] L. H. Boulogne, B. J. Wolf, M. A. Wiering, and S. M. van Netten,
small unmanned aircraft systems equipped with an airflow sensor array,” “Performance of neural networks for localizing moving objects with
Bioinspiration Biomimetics, vol. 9, no. 4, Nov. 2014, Art. no. 046015, an artificial lateral line,” Bioinspiration Biomimetics, vol. 12, no. 5,
doi: 10.1088/1748-3182/9/4/046015. Sep. 2017, Art. no. 056009, doi: 10.1088/1748-3190/aa7fcb.
[4] H. Fei, R. Zhu, Z. Zhou, and J. Wang, “Aircraft flight parameter detection [25] S. Callegari et al., “Experiments on aircraft flight parameter detection by
based on a neural network using multiple hot-film flow speed sensors,” on-skin sensors,” Sens. Actuators A, Phys., vols. 130–131, pp. 155–165,
Smart Mater. Struct., vol. 16, no. 4, pp. 1239–1245, Aug. 2007, doi: Aug. 2006, doi: 10.1016/j.sna.2005.12.026.
10.1088/0964-1726/16/4/035. [26] M. Schwerter, M. Gäding, M. Leester-Schädel, and A. Dietzel, “A flex-
[5] Y. Lian, W. Shyy, D. Viieru, and B. Zhang, “Membrane wing aerody- ible sensor system in foil for combined absolute pressure and flow
namics for micro air vehicles,” Prog. Aerosp. Sci., vol. 39, nos. 6–7, velocity measurements,” in Proc. 19th Int. Conf. Solid-State Sensors,
pp. 425–465, Aug./Oct.2003, doi: 10.1016/S0376-0421(03)00076-9. Actuat. Microsystems (TRANSDUCERS), Jun. 2017, pp. 1104–1107, doi:
[6] J. C. Ellsworth and S. A. Whitmore, “Simulation of a flush air-data 10.1109/TRANSDUCERS.2017.7994245.
system for transatmospheric vehicles,” J. Spacecraft Rockets, vol. 45,
no. 4, pp. 716–732, Jul. 2008, doi: 10.2514/1.33541.
[7] S. A. Whitmore, T. R. Moes, and T. J. Larson, “High angle-of-attack
flush airdata sensing system,” J. Aircr., vol. 29, no. 5, pp. 915–919, Zihao Dong received the B.S. degree in engineering from Beihang
Sep. 1992, doi: 10.2514/3.46262. University, Beijing, China, in 2017, where he is currently pursuing the
Ph.D. degree in mechanical engineering with the School of Mechanical
[8] S. A. Whitmore, J. Heeg, T. J. Larson, L. J. Ehernberger, F. W. Hagen, Engineering and Automation.
and R. V. Deleo. (1987). High-Angle-of-Attack Pneumatic Lag and His research interests include flexible electronics and sensor fusion.
Upwash Corrections for a Hemispherical Flow Direction Sensor.
[Online]. Available: http://hdl.handle.net/2060/19870014183
[9] S. Whitmore, T. Moes, M. Czerniejewski, and D. Nichols, “Application
of a flush airdata sensing system to a wing leading edge(LE- Zheng Gong received the B.S. degree in mechanical engineering from
FADS),” in Proc. 31st Aerosp. Sci. Meeting, Jan. 1993, pp. 1–18, doi: the Nanjing University of Aeronautics and Astronautics, Nanjing, China,
10.2514/6.1993-634. in 2018. He is currently pursuing the Ph.D. degree in mechanical
[10] S. Whitmore, B. Cobleigh, and E. Haering, “Design and calibra- engineering with the School of Mechanical Engineering and Automation,
tion of the X-33 flush airdata sensing (FADS) system,” in Proc. Beihang University, Beijing, China.
36th AIAA Aerosp. Sci. Meeting Exhibit, Jan. 1998, pp. 1–36, His research interests include bionic sensing, mechatronics, and
doi: 10.2514/6.1998-201. sensor fusion.
[11] R. Zhu, R. Que, and P. Liu, “Flexible micro flow sensor for micro aerial
vehicles,” Frontiers Mech. Eng., vol. 12, no. 4, pp. 539–545, Dec. 2017,
doi: 10.1007/s11465-017-0427-0. Bangqi Chen received the M.E. degree in engineering from Shanghai
[12] R. Zhu, P. Liu, X. D. Liu, F. X. Zhang, and Z. Y. Zhou, “A low- Maritime University, Shanghai, China, in 2023.
cost flexible hot-film sensor system for flow sensing and its application He is working as a part-time Research Assistant with Beihang Uni-
to aircraft,” in Proc. IEEE 22nd Int. Conf. Micro Electro Mech. Syst., versity, Beijing, China. His research interests include flexible electronics
Jan. 2009, pp. 527–530, doi: 10.1109/MEMSYS.2009.4805435. and sensor fusion.
[13] Z. Bin, T. Jianwu, Y. Fei, and Q. Lilin, “A new airspeed estima-
tion method for supersonic flow,” in Proc. IEEE Int. Conf. Signal
Process., Commun. Comput. (ICSPCC), Aug. 2013, pp. 1–4, doi:
10.1109/ICSPCC.2013.6663889. Tianyu Sheng is currently pursuing the Ph.D. degree in mechanical
[14] Y. Liu, D. Xiao, and Y. Lu, “Comparative study on a solving model and engineering with Beihang University, Beijing, China.
algorithm for a flush air data sensing system,” Sensors, vol. 14, no. 5, His research interests include piezoelectric flexible electronic integra-
pp. 9210–9226, May 2014, doi: 10.3390/s140509210. tion technology.
[15] C.-J. Zheng, Y.-P. Lu, and Z. He, “Improved algorithms for flush
airdata sensing system,” Chin. J. Aeronaut., vol. 19, no. 4, pp. 334–339,
Nov. 2006, doi: 10.1016/S1000-9361(11)60337-3. Yudong Cao received the B.S. degree in engineering from Beihang
[16] P. Pang et al., “Flexible skin for measurement of boundary layer state University, Beijing, China, in 2021, where he is currently pursuing the
and flight attitude identification on UAV,” Smart Mater. Struct., vol. 32, Ph.D. degree in mechanical engineering with the School of Mechanical
no. 4, Apr. 2023, Art. no. 045008, doi: 10.1088/1361-665X/acb4ca. Engineering and Automation.
[17] X. Na, Z. Gong, Z. Dong, D. Shen, D. Zhang, and Y. Jiang, “Flexible His research interests include bionic sensing, mechatronics, and
skin for flight parameter estimation based on pressure and velocity data sensor fusion.
fusion,” Adv. Intell. Syst., vol. 4, no. 6, Jun. 2022, Art. no. 2100276,
doi: 10.1002/aisy.202100276.
[18] W. Xiong et al., “Bio-inspired, intelligent flexible sensing skin for Qipei He received the B.S. degree in engineering from Southwest Jiao-
multifunctional flying perception,” Nano Energy, vol. 90, Dec. 2021, tong University, Chengdu, China, in 2018. He is currently pursuing the
Art. no. 106550, doi: 10.1016/j.nanoen.2021.106550. Ph.D. degree in mechanical engineering with the School of Mechanical
[19] Z. Dong, D. Shen, Z. Gong, D. Zhang, and Y. Jiang, “A flexi- Engineering and Automation, Beihang University, Beijing, China.
ble flow sensor for detecting airflow velocity and incident angle,” His research interests include flexible electronics and composite
IEEE Sensors J., vol. 22, no. 16, pp. 15884–15892, Aug. 2022, doi: nanomaterials.
10.1109/JSEN.2022.3190472.
[20] W. Xiong, D. Guo, Z. Yang, C. Zhu, and Y. Huang, “Conformable,
programmable and step-linear sensor array for large-range wind pressure
measurement on curved surface,” Sci. China Technol. Sci., vol. 63, Peng Zhao received the Ph.D. degree in engineering from Beihang
no. 10, pp. 2073–2081, Oct. 2020, doi: 10.1007/s11431-020-1642-4. University, Beijing, China, in 2022.
[21] Z. Dong et al., “Microfabrication of functional polyimide films and He is with the Beijing Institute of Space Launch Technology, Beijing.
microstructures for flexible MEMS applications,” Microsyst. Nanoeng.,
vol. 9, no. 1, p. 31, Mar. 2023, doi: 10.1038/s41378-023-00503-5.
[22] P. Zhao, P. Wu, D. Zhang, X. Ding, and Y. Jiang, “Cantilever-based Yonggang Jiang received the B.S. degree in mechanical engineering
differential pressure sensor with a bio-inspired bristled configuration,” from Beihang University, Beijing, China, in 2003, and the M.S. and Ph.D.
Bioinspiration Biomimetics, vol. 16, no. 5, Sep. 2021, Art. no. 055011, degrees in nanomechanics from the Department of Nanomechanics,
doi: 10.1088/1748-3190/ac1919. Tohoku University, Sendai, Japan, in 2006 and 2009, respectively.
[23] Y. Jiang et al., “Underwater source localization using an artificial lateral He is currently a Professor with the School of Mechanical
line system with pressure and flow velocity sensor fusion,” IEEE/ASME Engineering and Automation, Beihang University. His research inter-
Trans. Mechatronics, vol. 27, no. 1, pp. 245–255, Feb. 2022, doi: ests include nano-/micromanufacturing, bionic sensing, and micro-
10.1109/TMECH.2021.3062869. electromechanical systems (MEMS).

Authorized licensed use limited to: NANJING UNIVERSITY OF AERONAUTICS AND ASTRONAUTICS. Downloaded on May 13,2025 at 13:53:01 UTC from IEEE Xplore. Restrictions apply.