A Flexible Ultra-Thin Ultrasonic Transducer for Ice Detection on Curved Surfaces

Abstract

Icing phenomena occur on aircraft and unmanned aerial vehicles (UAVs) under extreme weather conditions. Ultrasonic detection technology is an effective method for measuring ice formation while maintaining the shape of the structure. However, current ultrasonic sensors, which are large and inflexible, are unsuitable for irregular UAV bodies, limiting their applications in real scenarios. For the detection of icing on curved structure, this study proposes a novel flexible ultra-thin ultrasonic transducer (FUTUT). The transducer exhibits excellent flexibility, making it suitable for use on high-curvature wings. Firstly, the FUTUT was designed based on the material properties of the airframe and the sensitivity requirements for ice detection, following the design guidelines for 1-3 type piezocomposites. The fabrication process for the FUTUT was then investigated, and its flexibility and low-temperature resistance were tested. Finally, icing detection experiments were conducted in an icing wind tunnel (IWT), where the FUTUT of 9.82 MHz demonstrated an ice-thickness-detection sensitivity of 0.29 mm. Experimental results indicate that the FUTUT possesses superior flexibility and exhibits excellent stability in low-temperature environments. These results underscore the FUTUT’s promise for applications in ice detection on curved structures.

1. Introduction

In practical scenarios, surfaces such as UAVs (unmanned aerial vehicles), aircrafts, bridges, cables, and roads are all susceptible to freezing. For aircraft, icing not only increases fuel consumption but also poses significant risks to flight safety [1]. While modern aircraft are equipped with Ice Protection Systems (IPSs) to mitigate icing, these systems often operate over large areas and extended durations under uncertain icing conditions, leading to substantial energy consumption. Ice detectors, capable of sensing the presence and thickness of ice accumulation, provide critical input for optimizing IPSs. Thus, the development of effective ice detection technology is essential for reducing energy consumption and enhancing flight safety.

Current ice detection methods primarily include vibration, optical, temperature, electrical, and ultrasonic techniques [2,3,4,5,6,7]. These diverse approaches allow for tailored solutions to meet specific application requirements. For instance, the vibration method rapidly detects ice formation by monitoring changes in resonant frequency but is unsuitable for quantitative measurement. Each method has inherent advantages and limitations, highlighting the need for ongoing research to optimize existing technologies for varied measurement needs. For example, capacitive ice detection sensors initially employed brittle metal probes [8], which have since been improved with graphene-based wafer sensing technologies [9]. Similarly, Love-wave ice sensors have evolved from rudimentary devices to thinner and more precise sensing systems [10,11,12,13,14].

Ultrasonic detection techniques (UDT), as a non-invasive measurement approach, also require advancements. Hansman first demonstrated the feasibility of planar ultrasonic transducers (UT) for ice thickness detection as early as 1988 [15,16]. Subsequent studies explored broader applications: Liu Y investigated water film characteristics and qualitative ice type identification [17,18,19], while Wang Y analyzed ice roughness and the freezing process of supercooled water films [20,21]. These studies underscore UDT’s significant potential in icing detection. More recently, Liu Q developed thin, flexible ultrasonic sensors using sol–gel spray technology, enabling conformal sensing on curved surfaces such as aircraft wings [22]. Similarly, Fuleki employed flexible ultrasonic sensors to monitor ice formation on a full-scale ALF502 turbofan engine [6,7]. Beyond aircraft wings and engines, platforms such as unmanned aerial vehicles (UAVs) [23], helicopters [24,25], and wind turbine blades also require effective ice detection solutions [26,27]. Compared to conventional sensors, the thin, flexible ultrasonic transducers offer distinct advantages for practical applications. Their conformal measurement capability minimizes structural modifications and integrates seamlessly with platforms like aircraft and UAVs, making them particularly suitable for advanced ice detection scenarios.

In this study, we propose a novel flexible ultra-thin ultrasonic transducer (FUTUT) for ice detection. Unlike the thin flexible sensors developed by Liu Q through spray deposition, the FUTUT does not require integration onto a substrate and can function independently. The sensor is specifically designed for aluminum skin and low-temperature environments, maintaining excellent ultrasonic signals even under such conditions. Furthermore, the FUTUT can be attached to high-curvature regions, such as the leading-edge stagnation point of a wing, allowing full coverage of the wing’s surface. With a sensitivity of up to 0.29 mm, the FUTUT satisfies the icing detection requirements outlined in aerospace standards [28].

2. Methods

2.1. The Principle and Design of the FUTUT

The structure of an ultrasonic transducer typically consists of a piezoelectric layer, a matching layer, and a backing layer, with the piezoelectric element being the core component of the transducer. Conventional ultrasonic transducers commonly use brittle piezoelectric ceramics for the piezoelectric layer, which cannot be bent as it lacks flexibility. Furthermore, piezoelectric ceramics such as PZT-5H have an acoustic impedance of approximately 34 MRayl, which is significantly higher than the impedance of aluminium aircraft skin (17 MRayl). This large mismatch in acoustic impedance can result in poor acoustic coupling between the transducer and the skin. As a consequence, a substantial portion of the piezoelectric acoustic energy fails to penetrate the skin, leading to significant elongation of the ultrasonic pulse and exacerbating signal oscillation. The role of the matching layer is to compensate for the acoustic impedance mismatch between the piezoelectric material and the tested object, which is why the matching layer is indispensable in conventional transducers.

This study employs 1-3 piezocomposite configuration as the piezoelectric layer, wherein piezocomposites are fabricated by cutting piezoelectric ceramics and filling them with epoxy resin [29,30,31]. The properties of the piezocomposite can be adjusted by varying the mixing ratio of piezoelectric ceramics and epoxy resin, allowing for the tailoring of acoustic impedance to match that of aluminum aircraft skin. This eliminates the need for a matching layer, thereby simplifying the structure of the ultrasonic transducer.

The FUTUT is developed using a 1-3 piezocomposite configuration. PZT-5H piezoelectric ceramics (3203HD, CTS Corporation, Bolingbrook, IL, USA) were selected as the active material due to their excellent piezoelectric and dielectric properties. Epoxy resin 301 was chosen as the passive material for its superior low-temperature resistance and stability.

Following the procedure in Refs. [32,33], we used all the assumptions presented in the study to derive the effective properties of the 1-3 piezoelectric composites.

$${\overline{c}}_{33}^E=V\left[c_{33}^E-{\displaystyle \frac{2V^{\prime }{(c_{33}^E-c_{12})}^2}{V(c_{11}+c_{12})+V^{\prime }(c_{11}^E+c_{12}^E)}}\right]+V^{\prime }{c}_{11}$$

(1)

$${\overline{e}}_{33}^E=[e_{33}-{\displaystyle \frac{2V^{\prime }{e}_{31}(c_{12}^E-c_{12})}{V(c_{11}+c_{12})+V^{\prime }(c_{11}^E+c_{12}^E)}}]$$

(2)

$${\overline{\epsilon }}_{33}^S=V[{\epsilon }_{33}^S+{\displaystyle \frac{2V^{\prime }{(e_{31})}^2}{V(c_{11}+c_{12})+V^{\prime }(c_{11}^E+c_{12}^E)}}]+V^{\prime }{\epsilon }_{11}$$

(3)

$${\overline{c}}_{33}^D={\overline{c}}_{33}^E+{\overline{e}}_{33}^2/{\overline{\epsilon }}_{33}^S$$

(4)

$$\overline{\rho }=V{\rho }^c+V^{\prime }{\rho }^p$$

(5)

In the above equations, the upper bar represents the composite material, and c_ij, e_ij, and e_ij, denote the elastic stiffness, piezoelectric constant, and dielectric constant, respectively. Superscripts E and S refer to the quantities at a constant electric field and strain, respectively. V and V’ are the volume percentages of the ceramic and epoxy resin, respectively, where V = 1 − V’. ${\rho }^c$ and ${\rho }^p$ are the densities of the ceramic and epoxy resin, respectively.

The required parameters of the 1-3 piezocomposites can be derived from the equations above. The acoustic impedance and frequency of the piezocomposites were obtained using the following equations:

$$\overline{Z}={({\overline{c}}_{33}^D\overline{\rho })}^{1/2}$$

(6)

$$f_c={({\overline{c}}_{33}^D/\overline{\rho })}^{1/2}/(2t)$$

(7)

where $\overline{Z}$ is the acoustic impedance of the 1-3 piezocomposite, the closer this value is to that of the measured material, the better the acoustic matching with the material; t is the thickness of the piezocomposite and f_c is the resonance frequency.

Using the effective medium model (EMM), the acoustic impedance and center frequency of 1–3 piezocomposite can be calculated as a function of volume fraction, as shown in Figure 1. Figure 1a shows the variation in acoustic impedance of the 1-3 piezocomposite with different piezoelectric phase volume fractions. Figure 1b illustrates the change in the center frequency with volume fraction. Since the center frequency varies with material thickness, the change in center frequency of the 1-3 piezocomposite with the piezoelectric phase volume fraction is provided for different material thicknesses. The parameters involved in the derivation of the properties of 1-3 piezocomposites are shown in

Table 1. The material properties of piezocomposite.

	PZT 5H	Epoxy	Piezocomposite
$C_{11}^E$ (1010 N/m2)	13.7	0.53	-
$C_{12}^E$ (1010 N/m2)	8.8	0.31	-
$C_{13}^E$ (1010 N/m2)	9.23	-	-
$C_{33}^E$ (1010 N/m2)	12.6	-	3.75
$e_{31}$ (C/m2)	−9.4	-	-
$e_{33}$ (C/m2)	22.5	-	17.73
${\epsilon }_{33}^S$/${\epsilon }_0$	1200.2	-	-
$\rho$ (103 kg/m3)	7.84	1.1	5.14
kt (%)	55	-	62.4

.

In Figure 1a,b, the dashed lines represent the acoustic impedance and center frequency values of the 1-3 piezocomposite when the piezoelectric phase volume fraction is 60%. And the center frequency can be determined to be 10.4 MHz when the thickness is 0.17 mm. The acoustic impedance of the 1-3 piezocomposite was found to be 18 MRayl, which closely matches the acoustic impedance of aluminum. Hence, when designing the FUTUT, the addition of a matching layer is unnecessary.

Based on the propagation characteristics of sound waves in media with different acoustic impedances, the sound waves emitted by the 1-3 piezocomposite will undergo reflection and transmission at the interface. The reflection and transmission coefficients are calculated using Equations (8) and (9):

$$r={\displaystyle \frac{R_{1\text{-}3}-R_{Al}}{R_{1\text{-}3}+R_{Al}}}$$

(8)

$$t={\displaystyle \frac{2R_{Al}}{R_{Al}+R_{1\text{-}3}}}$$

(9)

where R_1-3 = 18 MRayl and R_Al = 17 MRayl are the acoustic impedance for 1-3 piezocomposite and aluminum, respectively. Theoretically, the acoustic wave energy emitted by the 1-3 piezocomposite can efficiently penetrate the aluminum skin.

For icing detection on aircraft wing surfaces, the design of 1-3 piezocomposite focuses on a critical parameter: center frequency (f_c). Aerospace standards require that sensors detect ice thickness with a sensitivity of at least 0.3 mm [28]. To meet this requirement, a target center frequency of 10 MHz was selected. Based on the relationship between axial resolution and wavelength described in Ref. [34], a theoretical thickness resolution of 0.2 mm can be achieved at 10 MHz, which is sufficient for aircraft icing detection. As shown in Figure 1, for a piezocomposite material with a thickness of 0.17 mm and a piezoelectric phase volume fraction of 60%, the center frequency is determined to be 10.4 MHz, with an acoustic impedance of 18 MRayl.

Electromechanical and vibrational characteristics of the 1-3 piezocomposite were simulated using PZFlex 2017 (Weidlinger Associates Inc, Mountain View, CA, USA) finite element software, with the schematic diagram of the model shown in Figure 2.

In the model, the material and dimensional parameters of the 1-3 piezocomposite are consistent with those in the theoretical section. The width of the piezoelectric ceramic pillars is 80 µm, and the width of the epoxy resin is 20 µm. The frequency sweep range is from 0 to 20 MHz, and the resulting curve of electrical impedance-phase angle is shown in Figure 3. The calculated resonance frequency fr is 9.45 MHz, and the anti-resonance frequency fa is 9.51 MHz. Based on equation $f_c={\displaystyle \frac{f_r+f_a}{2}}$, the center frequency of the 1-3 piezocomposite is calculated to be 9.48 MHz.

2.2. Fabrication of the FUTUT

According to the theoretical results of the above-mentioned piezocomposite, the experimental fabrication was studied. Initially, the thickness of the piezoelectric ceramic was polished using a precision grinding machine (EJ-380IN, Engis, Wheeling, IL, USA), which ensures a flatness tolerance of ±1 μm for a 20 mm × 20 mm piezoelectric material. The material was first ground to a thickness of 300 μm, and then was cut using a high-precision dicing machine (Disco HI-TEC CHINA CO., LTD., Shanghai, China), with a blade (ZH05) width of 20 μm. Subsequently, epoxy resin was injected into the cut gaps and allowed to cure for 48 h. The 1-3 type piezocomposites were then ground to the target thickness, with the thickness measured using a precision thickness gauge (Sylvac_s, Dantsin, Switzerland), which provides a measurement accuracy of 0.1 μm. Followed by the deposition of gold electrodes on both the upper and lower surfaces using a magnetron sputtering system (FE-500, Shanghai Fuyi Vacuum Equipment Co., Ltd., Shanghai, China), the entire preparation process had been completed. The final thickness values at the center and four edges of the 1-3 piezocomposite material were 171.0 μm, 171.2 μm, 171.1 μm, 171.0 μm, and 170.8 μm, respectively. We could obtain an average thickness of 1-3 piezocomposite of 171 μm. The primary fabrication process is illustrated in Figure 4.

Furthermore, based on model coupling theory, the ratio of the width of the ceramic column (W₁) to the thickness of the piezocomposite should be less than 0.5 [31]. Accordingly, the dimensions were optimized to set W₁ = 80 μm and the cutting gap width (epoxy resin, W₂) to 20 μm. The structure of the piezocomposite was analyzed using an optical microscope (Olympus DSX1000, Tokyo, Japan). The dimensions of the 1-3 piezocomposite materials measured under the microscope were obtained by performing multiple measurements on a single sample and averaging the results. Figure 5a presents the surface imaging of the piezocomposite material, revealing measured dimensions of W₁ = 80 μm and W₂ = 20 μm, resulting in a calculated volume fraction of Ve = 64%. Figure 5b displays the cross-sectional imaging of the piezocomposite, where the measured thickness of the piezocomposite is t = 171 μm.

From the imaging of the 1-3 piezocomposite, it is clear that the piezoelectric ceramic columns and epoxy resin are well-formed, with no significant damage observed during the cutting and processing.

Based on the requirements of the application scenario, the transducers can be cut to the desired dimensions using a blade cutter. The prepared FUTUT samples are shown in Figure 6, with dimensions of 5 mm × 5 mm × 0.17 mm.

Electrical parameters of the piezocomposite were measured using an impedance analyzer (Agilent microtest 6632, San Jose, CA, USA), with the impedance-phase results presented in Figure 7. The resonance frequency (f_r) was measured at 8.45 MHz and the anti-resonance frequency (f_a) at 11.18 MHz, resulting in a calculated center frequency (f_c) of 9.82 MHz.

The comparison between the theoretical and experimental properties of the piezocomposite is summarized in

Table 2. Theoretical and experimental comparison of piezoelectric composites.

	t (μm)	Ve (%)	fc (MHz)	Z (MRayl)
Theory	170	60	10.4	18
Experiment	171	64	9.82	20.72

. The experimental results are largely consistent with the theoretical predictions, demonstrating the reliability of the fabrication process and material design.

2.3. Performance Testing of the FUTUT

(a) Bending performance

Before conducting the ice detection experiment, it was necessary to verify the bending performance of the FUTUT. The FUTUTs were affixed to cylindrical surfaces with different curvatures, and the interface between the piezoelectric phase and the epoxy resin phase inside the FUTUT was observed under an optical microscope to check for any delamination. Tests were performed at curvatures of diameters 90 mm, 28 mm, and 20 mm, with side-view imaging of the FUTUT samples shown Figure 8.

The optical microscope observations revealed that, under the curvature condition D₃ = 20 mm, no cracks or delamination occurred, thus confirming that the FUTUT exhibits flexible functionality from a morphological perspective. Despite no visible damage to the FUTUT’s appearance during bending, to ensure signal stability during use, the ultrasonic pulse echo signals and amplitude fluctuations in the bent state were further tested. The FUTUT was affixed to both a flat aluminum plate and the leading edge of a NACA 0024 wing, and the pulse echo signals were measured using an ultrasonic detection system, as shown in Figure 9.

Under bending conditions, the signal of the FUTUT exhibited significant changes. The peak of the first echo on the wing was higher than that on the flat plate, attributed to the FUTUT being attached to the inner surface of the wing, where the structural convexity at the leading edge resulted in a converging effect that amplified the sound wave energy compared to the flat surface. Phase differences in other echoes were primarily due to variations in the thickness of the wing and the flat plate. Although there were notable changes in the peak values of the pulse echo signals, these did not affect the functionality of the FUTUT on the wing.

(b) Resistance to high and low temperatures

After validating the FUTUT’s performance under bending conditions, low-temperature durability tests were conducted. The experiment was performed in a high-low temperature environmental test chamber to assess whether the FUTUT could operate in the temperature range of −55 °C to 65 °C, and to measure any shifts in the FUTUT’s center frequency and peak fluctuations under these conditions. Temperature intervals of 20 °C were used, and a total of seven sets of tests were conducted. The center frequencies and peak echo signal values of the FUTUT at different temperatures are shown in Figure 10.

During the low temperature chamber tests, after the temperature reached the target value, we waited for 30 min before measuring the center frequency and signal amplitude fluctuation of the FUTUT. Under the temperature range of −55 °C to 65 °C, the FUTUT exhibits a maximum center frequency shift of 0.23 MHz, corresponding to a relative deviation of only 0.23%. The maximum peak signal fluctuation is 0.043 mV, with a relative amplitude variation of merely 1.8%. Based on the test results of the center frequencies and signal peaks at varying temperatures, it can be concluded that the FUTUT designed in this study can withstand extreme temperatures ranging from −55 °C to 65 °C, maintaining good stability within this temperature range.

3. Ice Detection Experiments

A piezocomposite with dimensions of 5 mm × 5 mm × 0.17 mm was bonded to the leading edge of a NACA 0024 airfoil, as shown in Figure 11a. To ensure reliability under low-temperature and vibration conditions during the icing wind tunnel (IWT) test, a silver paste film was applied to the surface of the bonded piezocomposite, completing the fabrication of the FUTUT. The detailed structure of the FUTUT is illustrated in Figure 11b.

The performance of FUTUT was evaluated prior to the icing detection experiment. The pulse-echo response and its spectrum, measured on the wing, are shown in Figure 12. The FUTUT demonstrated a clear and distinct pulse-echo signal, confirming its capability to meet the requirements for ice thickness detection.

Icing detection experiments were conducted in the IWT at Nanjing University of Aeronautics and Astronautics. The test section measured 0.5 m in length, 0.3 m in width, and 0.4 m in height. Four freezing temperatures were selected for the study: −6.7 °C, −10 °C, −11 °C, and −12 °C. These conditions encompass the temperature ranges for glaze ice, mixed ice, and rime ice. The experiments were performed under constant environmental conditions, with only the temperature being varied. Wind speed was maintained at 60 m/s, liquid water content (LWC) at 0.8 g/m³, and median volume diameter (MVD) at 25 μm.

The icing detection system is illustrated in Figure 13. The FUTUT generates and receives signals through the voltage excitation of the DPR_300 pulse generator (Imaginant Inc., Pittsford, NY, USA) with a bandwidth of 60 MHz. The ultrasonic signal is collected and processed using a high-frequency data acquisition card with a sampling frequency of 250 MHz, while the pulse-echo signals are displayed in real time on a monitor through signal digital processor. A PT-100 platinum resistance thermometer is placed next to the FUTUT to monitor temperature variations during the experiment. To ensure continuous monitoring of the process of icing accumulation, the system remained operational throughout the experiments. The process of icing was terminated when the echo peak of the detected ice decayed to the point of being indistinguishable.

By comparing the pulse echo signals after icing with the initial non-icing signals, the ice echo can be clearly distinguished, as shown in Figure 14. A phase shift in the ice echo towards later times indicates the growth of the ice layer. The ultrasonic detection system used in this study has a sampling frequency of 20 Hz, allowing for real-time monitoring of the ice thickness growth process.

The ice thickness can be obtained by extracting the time difference in echo peaks and the propagation speed of sound waves, as shown in Equation (10):

$$h={\displaystyle \frac{1}{2}}c\Delta t$$

(10)

where c is the propagation speed of sound waves in the medium, and Δt is the propagation time of sound waves in the medium.

To enhance the clarity of the echo peak, envelope extraction was applied to the original pulse-echo signal [35], as shown in Figure 15. By employing the envelope signal processing technique, the peak values of the pulse echo signals and their corresponding phases can be more effectively captured, facilitating the accurate determination of ice thickness.

4. Results and Discussion

The echo peaks corresponding to different ice thicknesses are presented in Figure 16. The results indicate that the peak value of the pulse-echo signal decreases as ice thickness increases, reflecting the attenuation of ultrasonic wave propagation in porous media such as ice. By processing the pulse-echo signals throughout the freezing process, the growth curves of ice thickness at different temperatures were obtained in Figure 17. At varying temperatures, the FUTUT demonstrated the ability to measure a minimum ice thickness of 0.29 mm and maximum ice thicknesses of 2.23 mm, 3.71 mm, 2.71 mm, and 3.09 mm, respectively.

The ice samples at different temperatures are shown in Figure 18, with the images taken at the moment when ultrasonic measurements ceased. The characteristics of the ice samples at various temperatures are as follows: (1) At an icing temperature of −6.7 °C, the ice surface is relatively smooth and transparent, with an unfrozen water film remaining on the surface; (2) At icing temperatures of −11 °C and −12 °C, the ice appears milky white, with numerous small rough features on the surface; (3) At an icing temperature of −10 °C, the texture of the ice lies between the two aforementioned conditions, with a relatively smooth surface, although noticeable rough features appear at the upper and lower edges.

After the icing process was terminated, the ice shape was captured using tracing method. The ice was melted using heated copper plates, and the ice-shaped outline was drawn by hand. The traced ice shapes at different temperatures were shown in Figure 19. A vernier caliper was used to measure the ice thickness at positions corresponding to the FUTUT measurements. The recorded ice thicknesses were 2.1 mm, 3.65 mm, 2.62 mm, and 3.01 mm at −6.7 °C, −10 °C, −11 °C, and −12 °C, respectively.

The comparison of ice thickness measurements obtained using the FUTUT and the tracing method is presented in

Table 3. Comparison of ice thickness measured by ultrasonic and tracing method.

Temperature (°C)	FUTUT (mm)	Tracing (mm)	Error (%)
−6.7	2.23	2.1	6.19
−10	3.71	3.65	1.64
−11	2.71	2.62	3.44
−12	3.09	3.01	2.66

. Results indicate that the ice thickness measured by FUTUT is slightly greater than that measured by the tracing method. This discrepancy arises from the tracing method being a contact-based measurement; the copper plate’s higher heat capacity causes partial melting of the ice upon contact. Despite this, the maximum measurement error of the FUTUT system does not exceed 6.19%, indicating its reliable accuracy.

During the experiment measuring ice thickness growth using FUTUT, the icing phenomena under different temperatures are analyzed as follows:

(1) Under different icing temperature conditions, the minimum ice thickness measurable by FUTUT was 0.29 mm, although the actual value should be smaller. As shown in Figure 16, the ice thickness began to be calculated only after the envelope signals of the ice and the substrate were fully separated. Although the ice thickness can still be determined when echo signals overlap, no further signal processing was performed once the 0.3 mm ice detection standard specified in the aviation manual was satisfied [28].

(2) In the icing wind tunnel, the ice is formed by spray-impact freezing. During the early stages of icing, the ice surface is relatively smooth, and the effect of surface roughness on ultrasonic signal attenuation is not considered, resulting in smaller echo peak attenuation. In later stages, the accumulation of ice crystal particles creates larger roughness features, leading to significant attenuation of the ultrasonic signal. This is evident in Figure 16, where the echo amplitude attenuation is small for Ice1–Ice5, but becomes severe after Ice5.

(3) During the experiment, the icing process was terminated before the complete attenuation of the echo peak to ensure that the FUTUT could measure the final ice thickness, which was then compared with the tracing method. As a result, the maximum ice thickness measured at different temperatures does not represent the true limiting value.

Based on Table 3, the ice thickness range measured by FUTUT is largest at −10 °C. The maximum ice thickness at different temperatures do not follow a consistent pattern. In addition to the premature termination of icing during the experiment, we provide the following explanations based on observations during the icing process:

(1) At −6.7 °C, the ice is not completely frozen, with a thin water film on the surface (as shown in Figure 18a). This water film causes severe oscillation of the ultrasonic signal, reducing the echo peak amplitude [21] and leading to an overestimation of the ice thickness measurement limit, causing early termination of the freezing process. Interestingly, when the ice fully refroze, the echo peak amplitude increased.

(2) At −11 °C and −12 °C, supercooled large water droplets impact the wing surface, quickly freezing and forming a rough surface (as shown in Figure 18b). This results in increased attenuation of the ultrasonic signal, leading to a smaller measured ice thickness range compared to −10 °C. The difference in maximum ice thickness at −11 °C and −12 °C is that the insufficient temperature precision in the IWT and human control over the icing process.

(3) At −10 °C, supercooled large water droplets impact the wing surface, forming a water film that spreads and then freezes rapidly. As a result, the wing leading edge remains smooth, without the influence of surface roughness or an unfrozen water film.

Based on previous studies on ice porosity, surface roughness, and unfrozen water films on the ice surface, we found that the impact of ice porosity on ultrasonic signals is minimal, primarily causing attenuation in signal amplitude. In contrast, surface roughness or irregular ice shapes lead to significant attenuation of high-frequency ultrasonic signals, resulting in a considerable reduction in the measurement range. The presence of a water film on the ice surface interferes with the ultrasonic signal’s ability to assess ice thickness growth, a phenomenon that is inevitable under higher temperature icing conditions.

5. Conclusions

In summary, a high-precision flexible ultra-thin ultrasonic transducer (FUTUT) was developed for measuring ice thickness on curved surfaces, such as aircraft wings and UAVs. The sensor features a compact size and lightweight design, demonstrating excellent flexibility and stability under low-temperature conditions through a series of tests. Under the bending condition with a curvature diameter of D = 20 mm, the sensor remains intact without any cracks. In the temperature range of −55 °C to 65 °C, the shift in the sensor’s center frequency is only 0.23%, and the fluctuation in the signal amplitude is just 1.8%. Its capability for ice thickness measurement was demonstrated through the icing detection experiments conducted on a wing leading edge in the IWT. It is important to note that no additional processing is required on the tested wing model for the FUTUT installation. The experimental results demonstrate that the FUTUT at a frequency of 9.82 MHz can achieve an ice-thickness-measurement accuracy of 0.29 mm. Additionally, the ultrasonic icing detection system enables real-time monitoring of ice thickness growth. These findings confirm the feasibility of using the FUTUT for ice thickness measurements on curved structures. This study highlights the potential of the FUTUT to enhance the practical application of icing detection in real scenarios.