Experimental Study on Damage Detection of CFRP-Steel Interface Based on PMN-PT Piezoelectric Single Crystal

1. Introduction

Steel structures are prone to crack initiation, corrosion, and local damage under the combined effects of cyclic loading, environmental corrosion, temperature variation, and material aging during long-term service, seriously affecting structural safety and service life [1,2]. Traditional strengthening methods, including bolted connections, steel bonding, and welding reinforcement, can improve structural bearing capacity to a certain extent but suffer from complex construction, increased self-weight, potential damage to the parent material, and poor durability, making them difficult to meet the requirements of modern engineering for efficient, low-damage, and long-lasting strengthening [3,4,5,6,7,8]. Carbon fiber-reinforced polymer (CFRP) has the advantages of light weight, high strength, corrosion resistance, and convenient construction, and has been widely used in steel structure strengthening projects, forming CFRP–steel composite systems[9]. The collaborative performance of such systems highly depends on the quality of the interface bond. Interface debonding, adhesive aging, void defects, and cracks or groove damage in CFRP significantly reduce interface strength, inducing CFRP peeling or complete debonding failure, which seriously threatens the strengthening effect and structural safety[10]. Therefore, developing efficient, accurate, and quantitative interface damage detection techniques is of great engineering value for ensuring the safe operation of CFRP-strengthened steel structures.

Structural health monitoring (SHM) technology enables early identification, localization, and quantitative assessment of structural damage and has become an important means to ensure the safety of major infrastructure[11]. Piezoelectric materials, with their electromechanical coupling effect, fast response, moderate cost, and easy integration, are among the most commonly used smart sensing materials in SHM. Currently, lead zirconate titanate (PZT) ceramics and polyvinylidene fluoride (PVDF) piezoelectric films are widely used in engineering: PZT ceramics have good electromechanical coupling performance but are brittle and limited in sensitivity, making it difficult to capture subtle damage signals[12]; PVDF films are flexible and easy to process but have low piezoelectric constants and weak electromechanical coupling effects, leading to insufficient accuracy in high-frequency detection[13,14]. In recent years, lead magnesium niobate–lead titanate (PMN–PT) piezoelectric single crystals have shown great potential in high-frequency sensing, micro-strain measurement, and high-precision non-destructive testing due to their ultra-high piezoelectric constants, excellent electromechanical coupling performance, and low dielectric loss, providing a new material choice for high-precision detection of micro-damage in composite structures[15,16,17].

Piezoelectric-based ultrasonic guided-wave detection technology, with its advantages of active excitation, long propagation distance, and high sensitivity to minor defects, is highly suitable for detecting interface and internal defects in thin plates and composite structures[18,19]. As a typical ultrasonic guided wave, Lamb waves are highly sensitive to interface debonding, cracks, and delamination in thin-plate structures. However, their dispersion characteristics and multi-mode coexistence lead to complex signals and difficult interpretation, requiring dispersion curve analysis to select the optimal excitation frequency to reduce dispersion effects and ensure propagation of a single dominant mode[20,21]. Existing studies on CFRP–steel interface debonding detection mostly focus on qualitative identification, with limited research on quantitative characterization of debonding length and geometric parameters of internal CFRP groove defects, and lack high-sensitivity detection schemes based on PMN–PT single crystals[22].

Aiming at the above research gaps, this paper carries out the following work: (1) Develop externally bonded high-sensitivity piezoelectric sensors based on PMN–PT single crystals and conduct insulation, waterproofing, and electrical performance tests; (2) Analyze the dispersion characteristics of CFRP and steel plates using Disperse software to determine the optimal excitation frequency; (3) Fabricate CFRP–steel specimens with different debonding lengths and CFRP groove defects with varying geometric parameters, and collect time-domain signals using a PMN–PT ultrasonic guided-wave active detection system; (4) Extract first-arrival wave amplitude features, construct a damage index, and establish quantitative relationships between the damage index and debonding length/groove geometric parameters; (5) Verify the ability and accuracy of PMN–PT sensing technology to identify interface debonding and internal CFRP defects. The research results can provide theoretical basis and technical support for early, high-precision, and quantitative detection of interface damage in CFRP-strengthened steel structures.

2. Sensing Principle of PMN-PT Piezoelectric Single Crystal

Based on the direct piezoelectric effect and piezoelectric equations, PMN–PT piezoelectric patches are polarized along the thickness direction (z-axis), with the planar directions (x, y axes) parallel to the patch surface.

The first type of piezoelectric equations are adopted in this study[23] :

$$\begin{array}{cc}{\epsilon }_{\lambda }=c_{\lambda \mu }^E{\sigma }_u+d_{j\lambda }{E}_j& (u,\lambda =1,2,\mathrm{3...6})\\ D_i=d_{iu}{\sigma }_u+{\mu }_{ij}^{\sigma }{E}_j& (i,j=1,2,3)\end{array}$$

(1)

The first equation describes the relationship between strain (${\mathsf{ϵ}}_{\mathsf{\lambda }}$), stress (${\mathsf{\alpha }}_{\mathrm{u}}$), and electric field (${\mathrm{E}}_{\mathrm{j}}$), reflecting the influence of external stress and electric field on strain. The second equation describes the relationship between electric displacement (${\mathrm{D}}_{\mathrm{i}}$), stress (${\mathsf{\alpha }}_{\mathrm{u}}$), and electric field (${\mathrm{E}}_{\mathrm{j}}$), reflecting the influence of electric field and stress on electric displacement.

Considering that the cross-sectional areas in the x and y directions are significantly smaller than that in the z direction, under sensing mode with no external electric field, the quantitative relationship between the output charge and electric displacement of the PMN–PT patch can be expressed as:

$$Q={\displaystyle \iiint \left[D_1{D}_2{D}_3\right]}\left\{\begin{array}{c}d{A}_1\\ d{A}_2\\ d{A}_3\end{array}\right\}$$

(2)

Since the lateral areas dA₁ and dA₂ are negligible compared with the main surface area dA₃, Equation (2) simplifies to:

$$Q=D_3{A}_3$$

(3)

Substituting Equation (1) into Equation (3) yields:

$$Q={\displaystyle \sum _{i=1}^3{d}_{3i}}{\sigma }_i{A}_3=\left(d_{31}{\sigma }_1+d_{32}{\sigma }_2+d_{33}{\sigma }_3\right)A_3$$

(4)

The charge Q is converted to voltage U by a DH-5922D dynamic acquisition instrument:

$$U={\displaystyle \frac{Q}{C}}$$

(5)

where C is the feedback capacitance of the measuring instrument. A DH-5922D high-performance dynamic signal acquisition system and an HK-9209 charge amplifier are used for signal processing.

3. Experimental Research on Damage Detection of CFRP-Steel Composite Structure Using PMN-PT Ultrasonic Guided Waves

3.1. Dispersion Characteristics Analysis of CFRP-Steel Plates

To investigate the propagation characteristics and dispersion relations of ultrasonic waves in CFRP–steel specimens, dispersion curve analysis is performed. The basic physical parameters of CFRP and steel plates are input (Table 1 and Table 2). The geometric model is set with a 2 mm thick CFRP plate, a 5 mm thick steel plate, and vacuum boundaries on the top and bottom surfaces. The calculation range is 0–600 kHz (upper limit), with a maximum phase velocity of 20 m/ms and a frequency step of 1 kHz[24].

The A0 mode (anti-symmetric) is dominated by bending vibration, with opposite particle motion directions on the top and bottom surfaces and energy concentrated near the surface. At low frequencies, A0 has a long wavelength and low velocity, increasing rapidly with frequency before stabilizing.

The S0 mode (symmetric) is dominated by axial tension–compression motion, with identical particle motion directions on the top and bottom surfaces and energy distributed uniformly across the thickness, propagating much faster than A0.

Dispersion curves (phase velocity and energy velocity) for CFRP and steel plates in 0–600 kHz are obtained (Figure 1). For CFRP, only A0 and S0 modes exist below 450 kHz; A1 mode appears above 450 kHz. A0 velocity varies significantly at low frequencies but stabilizes at high frequencies; S0 velocity is nearly constant from 0–400 kHz, indicating high propagation efficiency. Phase and energy velocities at 100–300 kHz are listed in Table 3.

For steel, four modes (A0, A1, S0, S1) exist from 0–600 kHz, with increasing mode numbers and stronger dispersion at higher frequencies. From 0–300 kHz, A0 phase and energy velocities increase rapidly then stabilize, with energy velocity > phase velocity; S0 velocity decreases slightly. From 300–600 kHz, A1 and S1 modes emerge: A1 velocity stabilizes at high frequencies with energy velocity < phase velocity; S1 phase velocity decreases rapidly then stabilizes, while energy velocity increases rapidly then stabilizes (Table 4).

Within 100–300 kHz, only A0 and S0 modes exist in both CFRP and steel, minimizing modal complexity. To reduce dispersion and ensure stable guided-wave propagation, 150 kHz is selected as the optimal excitation frequency.

3.2. Interface Debonding Detection Test of CFRP-Steel

3.2.1. Test Design

PMN–PT sensors are bonded symmetrically on both sides of the damage area in a pitch–catch configuration: one as an actuator, the other as a sensor (Figure 2). Specimens with debonding lengths of 0–40 mm (H1–H5) are fabricated (Table 5).

The wave-based method using PMN-PT piezoelectric single crystal sensing is an active detection technique for structural monitoring, which requires selecting an appropriate waveform as the system excitation source[25]. The signal frequency is set to 150 kHz.

In this experiment, a five-cycle sinusoidal wave modulated by a Hanning window with a frequency of 150 kHz is selected as the excitation signal, with a maximum output voltage of 20 V[26]. The waveform is shown in Figure 3, and its expression is as follows:

$$y(t)={\displaystyle \frac{A}{2}}\left(1-\cos \left({\displaystyle \frac{2\pi ft}{n}}\right)\right)\sin \left(2\pi ft\right),n=5,f=150\mathrm{kHz}$$

(6)

where t denotes time, A is the signal amplitude, f is the excitation frequency, and n is the number of peaks.

Instruments required for measurement are signal generator, digital oscilloscope, and signal amplifier. The details of the experimental setup are shown in Figure 4. The computer is connected to the signal generator; the modulated output waveform is transmitted to the PMN-PT sensor to excite guided waves propagating in the specimen. The signal received by the other PMN-PT sensor is amplified by 10 times, then transmitted to the oscilloscope for acquisition, display, and filtering for subsequent analysis. Each test case is sampled three times and averaged to reduce measurement errors and improve data accuracy.

The experimental testing system is shown in Figure 4. A five-cycle sinusoidal wave modulated by a Hanning window is adopted for signal testing. The parameters of the signal generator are set as follows: Channel 1 (CH1) outputs a five-cycle Hanning-windowed sinusoidal pulse wave with a frequency of 150 kHz and a peak voltage of 20 V. Channel 2 (CH2) outputs a Lorentz pulse wave at 100 Hz. The trigger mode is set to the MOD button, the trigger source is CH2, and the trigger count is set to 1. That is, a five-cycle modulated pulse is triggered every 10 ms.

This paper focuses on denoising using an FIR bandpass filter. By setting the frequencies of the stopband, passband, and transition band, the FIR filter allows signals within a specific frequency range to pass through while effectively suppressing out-of-band signals, which is particularly effective for removing unwanted high- and low-frequency noise.

In the experiment, the firpm function is used to design an optimized FIR filter. The transfer function coefficients and filter order are adjusted to achieve the optimal filtering performance. A comparison of signals before and after FIR filtering shows that the filter successfully removes frequency components outside the 100 kHz–200 kHz range, significantly reducing interference. This results in cleaner, clearer signals that facilitate subsequent accurate analysis and mode extraction. The voltage waveforms before and after filtering are presented in Figure 5. All voltage signals analyzed and processed in the subsequent sections refer to the filtered signals.

3.2.2. Waveform Signal Analysis Under Different Debonding Lengths

The voltage time-domain waveform of the undamaged condition H1 is obtained through testing the above experimental cases, as shown in Figure 6.

First, time-domain feature analysis is performed on the undamaged CFRP–steel specimen. Since structural boundaries significantly affect signal propagation, ultrasonic waves reflected by boundaries are received again by the PMN-PT sensor, interfering with the acquired time-domain signals. Thus, boundary reflection waves are considered, and the waveform is divided into three wave packets.

The signal waveform of the debonding-free condition (H0) is shown in Figure 6. Wave packet ① is mainly induced by the S0 mode ultrasonic wave, which propagates rapidly and arrives first at the PMN-PT sensor. According to the data, the arrival time is 18.4 μs. The excitation duration is 33.35 μs; between 18.4 μs and 51.7 μs, wave packet ① contains five peaks, consistent with the five-cycle Hanning-windowed excitation signal. However, due to the small size of the CFRP plate, sidewall reflections cause minor waveform differences from the excitation signal. The maximum amplitude of the S0 mode wave occurs at 42.6 μs, with a voltage of 73.3 mV.

Wave packets ② and ③ result from the superposition of reflected waves and A0 mode ultrasonic waves. The A0 mode propagates slowly in the limited-size specimen and overlaps with boundary reflections, making it difficult to identify clearly. The maximum amplitude of the superimposed reflected waves appears at 96.4 μs, with a voltage of 334 mV.

Ultrasonic testing was perfrmed on specimens under different debonding conditions, and the waveform diagrams for Conditions 2 to 5 are obtained, as shown in Figure 7.

It can be observed from the figure that the first-arrival time gradually decreases with the change of test conditions, from 18.4 μs for Condition H1 to 17.4 μs for Condition H5, a relative change of approximately 5.4%. This indicates that the arrival time of ultrasonic signals decreases as the degree of material or structural damage increases, which may be attributed to structural changes. However, this trend is slow and insignificant, making it difficult to accurately determine the specific debonding severity.

The time of the first-arrival amplitude fluctuates slightly across different conditions: 42.6 μs for H1, 42.8 μs for H2, 42.2 μs for H3, 44.2 μs for H4, and 43.2 μs for H5. Although the fluctuations are small, no clear or consistent trend can be identified. The voltage of the first-arrival amplitude gradually increases from 73.3 mV (H1) to 81.3 mV (H3), then becomes negative for H4 and H5, at −84.3 mV and −92.6 mV, respectively. This variation shows that the signal amplitude increases with damage severity; under more severe damage (H4 and H5), the amplitude becomes negative and the peak of the first-arrival wave shifts backward. Analyzing the first-arrival amplitude can effectively assess the degree of material or structural damage.

The time of the maximum reflected wave amplitude Increases gradually with the test conditions: 96.4 μs for H1, with an overall delay for H2 to H5, though no clear pattern is observed. The reflected wave amplitude varies significantly: 334 mV (H1), 343.8 mV (H2), −305 mV (H3), −390.9 mV (H4), and back to positive at 384 mV (H5). The irregular fluctuations indicate that the reflected wave behavior is complex, influenced by multiple factors including damage complexity and wave propagation interference. These changes suggest that reflected wave propagation becomes more complicated, possibly due to multiple reflections or altered propagation paths caused by damage, leading to inconsistent variations.

Based on the above analysis of key temporal features in the waveforms, the specific values of first-arrival time, first-arrival amplitude time, first-arrival amplitude voltage, maximum reflected wave amplitude time, and reflected wave amplitude voltage for each condition are summarized in Table 6. The table clearly and intuitively presents the variations of key temporal and amplitude features across all test conditions.

It can be seen from the table that the amplitude voltage of the first-arrival signal rises gradually as the damage becomes more severe. This reveals a strong correlation between the variation of the first-arrival amplitude voltage and the debonding degree of the CFRP-steel specimen. The finding provides a valuable reference for the health monitoring of CFRP-steel structures. Figure 8 presents the comparison of first-arrival signal waveforms under different conditions.

The above analysis enables qualitative identification of interfacial debonding damage in CFRP-steel specimens. To further quantitatively characterize the debonding severity, statistical analysis is conducted on the measured data in this study. A debonding damage index $D{I}_i$ based on the first-arrival signal amplitude acquired by PMN-PT sensors is defined as follows:

$$D{I}_{\beta }=\left|{\displaystyle \frac{V_i-V_0}{V_0}}\right|$$

(7)

where $i$ denotes different debonding lengths, $V_0$is the first-arrival amplitude voltage under the intact condition, and $V_i$represents the first-arrival amplitude voltage corresponding to each debonding length.

When $D{I}_{\beta }=0$, the CFRP-steel specimen is in a healthy state. As illustrated in Figure 9, the damage index is 0 for Condition H1, 4.4% for H2, 10.9% for H3, 15% for H4 and 26.3% for H5[27]. The damage index increases with the debonding length of the specimen, demonstrating that the amplitude of the first-arrival signal detected by the PMN-PT sensor also rises under corresponding test conditions[28].

3.3. Identification Test of Internal Groove Defects in CFRP Plates Using PMN-PT Ultrasonic Guided Waves

To investigate the performance of damaged CFRP plates, CFRP specimens with dimensions of 200 × 50 × 5 mm were adopted in this test, while the sizes of other specimens remained unchanged. The groove defects of the CFRP plate in the CFRP-steel composite specimen are presented in Figure 10. To effectively detect various groove defects in CFRP plates, five groups of test conditions were designed, as listed in Table 7.

Waveforms under working conditions M1 to M5 are obtained by analyzing different groove defects of CFRP, as shown in Figure 11.

It can be observed that the waveforms captured by the PMN-PT sensor differ remarkably among conditions M1 to M5 without an obvious overall trend. Further analysis on the first-arrival waveforms shows that the maximum amplitude is 69.5 mV for M1, 76.7 mV for M2, 82.2 mV for M3, 63.1 mV for M4 and 52.7 mV for M5.

With fixed groove depth and width, the maximum amplitude of the first-arrival signal increases gradually as the groove length rises for M1, M2 and M3 (Figure 12a). When the groove length and width remain unchanged, the maximum amplitude decreases steadily with the increase of groove depth for M2, M4 and M5 (Figure 12b).

Different groove defects in CFRP plates lead to distinct variation rules of the first-arrival signals. As the groove length increases, more ultrasonic energy is retained within the CFRP plate, resulting in a gradual rise of signal amplitude. In contrast, with the increase of groove depth, the propagation path for ultrasonic waves narrows. Part of the energy is reflected at the grooves, which reduces the amplitude of the received first-arrival signal.

In conclusion, 150 kHz is determined as the optimal excitation frequency. Frequency domain analysis further verifies that the dominant frequency of the received signals matches the excitation frequency, though the amplitude shows weak regular variation. The damage index defined based on the first-arrival amplitude increases monotonously with the growing debonding length, proving its excellent performance in damage evaluation. For internal groove defects of CFRP plates, test results indicate that the first-arrival amplitude rises with longer grooves at a fixed depth, and declines with deeper grooves at a fixed length. The experimental findings validate the reliability of this method, which provides a feasible technical approach for the detection and evaluation of internal defects in CFRP plates.

4. Conclusions

This study takes the CFRP-steel composite strengthening system as the research object. A high-sensitivity surface-mounted sensor is developed using PMN-PT piezoelectric single crystal material. Combined with the active ultrasonic guided wave detection technology, systematic research is conducted on the identification and quantitative evaluation of interface debonding damage and CFRP groove defects. The main conclusions are summarized as follows:

(1)The PMN-PT piezoelectric sensor exhibits superior performance and strong applicability.The developed surface-mounted PMN-PT piezoelectric sensor possesses good i ion, water resistance and electrical stability. Its capacitance variation conforms to theoretical rules, with the sensitivity error controlled within 4.7%. It can stably capture tiny damage signals in complex engineering environments, laying a solid sensing foundation for high-precision detection.

(2)The optimal excitation frequency is determined through dispersion characteristic analysis.Dispersion curves of CFRP plates and steel plates are analyzed via Disperse software, and 150 kHz is confirmed as the optimal excitation frequency. At this frequency, the Lamb wave presents a single mode with weak dispersion effect, which effectively suppresses interference from multi-mode signals and ensures stable guided wave propagation and interpretable signals.

(3)The amplitude of the first-arrival signal can effectively characterize the degree of interface debonding and realize quantitative evaluation.Specimens with debonding lengths ranging from 0 to 40 mm are fabricated in gradient. Time-domain analysis of ultrasonic signals shows that the amplitude of the first-arrival signal varies monotonously with the increase of debonding length. The damage index (DI) established based on the first-arrival amplitude has a good linear correlation with debonding length, enabling accurate quantification of interface debonding damage.

(4)The amplitude of the first-arrival signal is sensitive to geometric parameters of CFRP groove defects.Tests are carried out on CFRP groove defects with different lengths and depths. The results reveal that the first-arrival amplitude increases with the growth of groove length at a fixed depth, and decreases with the increase of groove depth at a fixed length. This rule can effectively distinguish geometric differences of internal defects in CFRP and provide a quantitative reference for material defect detection.

(5)The PMN-PT ultrasonic guided wave detection technology features high reliability and great application potential.Experimental results verify that the combination of PMN-PT piezoelectric sensing and ultrasonic guided wave technology can identify CFRP-steel interface debonding and internal CFRP groove defects with high sensitivity, and realize qualitative identification and quantitative evaluation of damages. With low implementation cost, high detection accuracy and wide adaptability, this technology serves as an efficient and reliable solution for structural health monitoring of steel structures strengthened with CFRP.