A method for detecting backside defects in a composite material

Abstract

The present invention belongs to the field of non-destructive testing, and discloses a device and method for detecting defects on the back surface of a composite material. The device includes: a probe, a signal acquisition module, and a signal processing module. The probe includes a transmitting coil and a plurality of receiving coils. The transmitting coil and the plurality of receiving coils are in the same plane and their axes are parallel. The distance between the transmitting coil and the receiving coils is the coil pitch. The coil pitch is set to maximize the effective mutual inductance change caused by the defect area on the back surface of the specimen while suppressing the mutual inductance coupling background field between the transmitting coil and the receiving coils. The present invention conducts a collaborative design of the LTR probe structure and the EMD modal reconstruction strategy: by optimizing the probe structure, the mutual inductance coupling interference between the transmitting coil and the receiving coils is reduced from a physical level; for the low-frequency lift-off noise that becomes more sensitive due to the increase in the coil pitch, an adaptive signal processing method based on empirical mode decomposition is introduced for effective suppression, improving the signal-to-noise ratio of the detection signal.

Description

Technical Field

[0001] This invention belongs to the field of nondestructive testing, and specifically relates to a device and method for detecting defects on the back side of composite materials. Background Technology

[0002] Eddy current testing is a rapid, non-contact, non-destructive testing method based on the principle of electromagnetic induction, and it has been widely used for defect detection in conductive components. Its basic principle is as follows: the alternating magnetic field applied by the probe's excitation coil induces eddy currents in the component; when a defect is present, the eddy current distribution is disturbed, thereby changing the local magnetic field and the induced voltage of the receiving coil. By analyzing the amplitude, real part, imaginary part, or phase of the output signal from the receiving coil, the defect can be identified.

[0003] However, applying this technology to composite materials, especially anisotropic materials such as CFRP, still faces significant challenges in detecting back-side defects. This is mainly due to two reasons: First, composite materials are highly nonlinear and heterogeneous, with strong electromagnetic attenuation, making it difficult for the excitation magnetic field to effectively penetrate deep into the layers, resulting in extremely weak response signals from back-side defects. Second, traditional narrow-pitch LTR probes (Lateral transmit-receive probes) suffer from strong direct mutual inductance coupling between their transmitting coil (T) and receiving coil (R), which masks weak perturbation components related to deep-layer defects in the received signal, leading to excessively strong background signals and severely limiting the detection signal-to-noise ratio. However, existing research typically considers coil structure design and signal processing algorithms separately, and a collaborative design approach has not yet been developed to reduce the difficulty of capturing back-side defect perturbations by increasing the coil spacing, thereby introducing low-frequency noise enhancement.

[0004] Among them, patent application CN113671022A discloses a lift-off measurement device and method based on the crossover point of the coil spacing of a pulsed eddy current detection probe. It obtains differential signal curves by adjusting different coil spacings and constructs a quantitative evaluation model of lift-off based on the time of the differential signal crossover point to achieve measurement of unknown lift-off. However, the adjustment of the coil spacing in this method is still limited to the structural parameter range of traditional probes. Its design purpose mainly serves lift-off compensation and measurement accuracy improvement, without optimizing the system from the perspective of increasing the spacing between the transmitting and receiving coils to reduce direct coupling and enhance the detectability of back-side defect disturbances. Another patent application, CN108152363A, discloses a pipeline defect identification method based on the inherent time-scale decomposition of the suppressed end. Although the background section mentions empirical mode decomposition for defect identification, it does not achieve the goal of making the receiving coil more easily capture back-side defect disturbances by combining coil spacing adjustments. Summary of the Invention

[0005] The purpose of this invention is to provide a composite material back defect detection device, which reduces the direct coupling interference between the transmitting coil and the receiving coil by increasing the coil spacing between the transmitting coil and the receiving coil in a traditional LTR probe, making it easier for the receiving coil to capture disturbances of back defects.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: a composite material backside defect detection device, comprising:

[0007] The probe includes a transmitting coil and several receiving coils. The transmitting coil and several receiving coils are in the same plane and their axes are parallel. The distance between the transmitting coil and the receiving coil is called the coil spacing, denoted as S. The coil spacing is set to maximize the effective mutual inductance change caused by the receiving coil to the defect area on the back of the specimen while suppressing the background field of mutual inductance coupling between the transmitting coil and the receiving coil.

[0008] A signal acquisition module is configured to acquire a detection signal output by the receiving coil that includes a defect signal and lift-off noise;

[0009] A signal processing module is configured to perform empirical mode decomposition on the detection signal to identify and suppress low-frequency lift-off noise associated with the coil spacing.

[0010] In another implementation, the signal processing module is configured to perform empirical mode decomposition on the detection signal, decomposing the detection signal into multiple intrinsic mode function components and a residual term; identifying the residual term as low-frequency trend noise caused by lift-off variation and removing it; and reconstructing the retained intrinsic mode function components to obtain a defect signal with suppressed lift-off noise.

[0011] In another embodiment, the value of the coil spacing is chosen to maximize the voltage change of the receiving coil. The voltage change is denoted as ΔV, which is the peak point of the S-ΔV relationship curve corresponding to the coil spacing, where the voltage change is the difference in detection voltage amplitude between defective and defect-free conditions.

[0012] In another embodiment, when the probe has multiple receiving coils, the arrangement of each receiving coil and the transmitting coil is as follows: a linear array arrangement, in which each receiving coil is arranged along a straight line and the distance between it and the transmitting coil increases sequentially; or a ring array arrangement, in which each receiving coil is distributed in one or more concentric circles with the transmitting coil as the center; or a matrix array arrangement, in which the transmitting coil and several receiving coils are distributed in a grid pattern in rows and columns.

[0013] The present invention also provides a detection method based on the above-mentioned composite material backside defect detection device, characterized in that it includes the following steps:

[0014] A1. Determine the coil spacing of the probe: Based on the ply structure of the test specimen and the expected depth and length of the back defect, adjust the coil spacing between the transmitting coil and the receiving coil of the probe to maximize the effective mutual inductance change caused by the receiving coil to the back defect area of ​​the test specimen.

[0015] A2. Signal Acquisition: The specimen is scanned using an LTR probe with the optimized coil spacing to acquire the raw eddy current detection signal;

[0016] A3. Adaptive filtering: Perform empirical mode decomposition on the original eddy current detection signal, adaptively decompose it into several intrinsic mode function components and a residual term, and based on the coil spacing parameter, adaptively identify and remove the low-frequency lift-off noise trend term related to the coil spacing in the decomposition result;

[0017] A4. Defect Signal Reconstruction and Evaluation: Defect signals are reconstructed using the selected intrinsic mode function components, and the detection and evaluation of defects on the back side of composite materials are achieved based on their amplitude, phase, or signal-to-noise ratio.

[0018] In another implementation, step A1 uses numerical simulation analysis to determine the coil spacing, which includes: within a preset coil spacing range S1 < S < S N N is a positive integer. Electromagnetic simulation models of the probe and composite material specimens under different coil spacing conditions are established respectively. The eddy current density distribution, magnetic field penetration depth and induced voltage of the receiving coil at different depths inside the composite material under different coil spacing conditions are analyzed. The amplitude and phase change trends of the detection signal under different coil spacing conditions are compared. Based on the above comparison results, the coil spacing range that achieves a balance between enhancing the response to back defects and reducing the mutual inductance coupling between coils is determined.

[0019] In another implementation, in step A2, the acquired eddy current detection raw signal includes the real part signal and the imaginary part signal of the voltage output by the receiving coil.

[0020] In another implementation, in step A3, the process of adaptively identifying and removing low-frequency lift-off noise trend terms includes: determining the last residual component that satisfies the monotonicity condition obtained from empirical mode decomposition as the main noise component related to lift-off changes; further analyzing the frequency characteristics of its preceding intrinsic mode function component, and if its energy is mainly concentrated in the extremely low frequency band determined by the probe scanning speed, then classifying it as lift-off related noise and removing it.

[0021] In step A3, the algorithm flow based on empirical mode decomposition includes:

[0022] a) Obtain the original eddy current detection signal h(t)=x(t);

[0023] b) Find all extreme points in x(t) and construct the upper and lower envelopes, and calculate the envelope mean m(t);

[0024] c) Calculate h k (t) = h(t) - m(t);

[0025] d) If h k If (t) satisfies the intrinsic mode function condition, then it is denoted as IMF. n (t), otherwise h k (t) is the new signal. Repeat steps (b)-(c);

[0026] e) Subtract the extracted intrinsic mode functions to obtain the residual r. n (t):

[0027]

[0028] f) If the residual r n If h(t) is not a monotonic function, then treat it as a new h(t) and repeat steps (b)-(e).

[0029] g) Identify low-frequency noise and separate it to the residual trend term r n In (t), removal effectively suppresses lift-off noise while preserving defect signal characteristics:

[0030]

[0031] The residual terms that satisfy the monotonicity condition are identified as the main trend term noise related to lift-off and removed during the reconstruction process. Subsequently, the retained intrinsic mode function (IMF) components are reconstructed to obtain the defect response signal after suppressing lift-off noise. This processing method can adaptively separate low-frequency disturbances and defect features without relying on preset filtering parameters, which is beneficial for improving the stability and signal-to-noise ratio of the detection signal. Frequency characteristic analysis is performed on the IMF component immediately preceding the residual term to determine whether its energy distribution is mainly concentrated in the extremely low frequency range determined by the probe scanning speed. When the spectral characteristics of this IMF component are dominated by quasi-DC or extremely low-frequency components, and its trend is consistent with the lift-off changes caused by probe scanning, this IMF component is identified as a lift-off related noise component and removed during signal reconstruction.

[0032] In another implementation, in step A4, the signal-to-noise ratio is calculated in the following manner to quantitatively evaluate the detection effect:

[0033]

[0034] Among them, V d V represents the voltage change of the receiving coil in the defect area.n This represents the voltage change of the receiving coil in a defect-free region.

[0035] The present invention also provides a defect signal processing method based on the above-mentioned composite material backside defect detection device, which includes the following steps:

[0036] B1. Determine the coil spacing of the probe: Based on the ply structure of the test specimen and the expected burial depth d and length L of the back defect, adjust the coil spacing between the transmitting coil and the receiving coil of the probe to maximize the effective mutual inductance change caused by the receiving coil to the back defect area of ​​the test specimen.

[0037] B2. Signal Acquisition: The specimen is scanned using an LTR probe with the optimized coil spacing to acquire the raw eddy current detection signal;

[0038] B3. Adaptive Filtering: Perform empirical mode decomposition on the original eddy current detection signal, adaptively decompose it into several intrinsic mode function components and a residual term, and based on the coil spacing parameter, adaptively identify and remove the low-frequency lift-off noise trend term related to the coil spacing in the decomposition result;

[0039] B4. Defect signal reconstruction: The defect signal is reconstructed using the filtered intrinsic mode function components.

[0040] The beneficial effects of this invention are as follows: This invention coordinates the design of the LTR probe structure and the modal reconstruction strategy: by optimizing the probe structure, the mutual inductance coupling interference between the transmitting coil and the receiving coil is reduced from a physical level; for the low-frequency lift-off noise that is more sensitive due to the increased coil spacing, an adaptive signal processing method based on empirical mode decomposition is introduced for effective suppression, thereby improving the signal-to-noise ratio of the detection signal; the coordinated design of the probe structure parameters and the signal processing method enables the detection device and method to be applicable to the detection of defects on the back side of composite materials with different burial depths and sizes, and has good stability and applicability. Attached Figure Description

[0041] Figure 1 A flowchart of the detection method using probes and modal reconstruction;

[0042] Figure 2a A side view showing the configuration of the transmitting and receiving coils and the flat-cut probe;

[0043] Figure 2b A three-dimensional diagram showing the configuration of the transmitting and receiving coils and the flat-cut probe;

[0044] Figure 3a A side view showing the configuration of the transmitting and receiving coils and the absolute value type probe;

[0045] Figure 3bA three-dimensional diagram showing the configuration of the transmitting and receiving coils and the absolute value type probe;

[0046] Figure 4a A side view showing the configuration of the transmitting and receiving coils and the LTR probe;

[0047] Figure 4b A three-dimensional diagram showing the configuration of the transmitting and receiving coils and the LTR probe.

[0048] Figure 5 This is a schematic diagram of the magnetic field lines distribution of an absolute value type probe;

[0049] Figure 6a This is a schematic diagram of the magnetic field distribution of a traditional narrow-pitch coil LTR probe.

[0050] Figure 6b For comparison Figure 6a A schematic diagram of the magnetic field distribution of an LTR probe with a larger coil spacing;

[0051] Figure 6c For comparison Figure 6b A schematic diagram of the magnetic field distribution of an LTR probe with a larger coil spacing;

[0052] Figure 7 To be at the same tilt angle Lower coil spacing (S1) S2) and lift-off distance (l1) The relationship diagram of l2);

[0053] Figure 8 This is a schematic diagram illustrating the effect of coil spacing on the lift-off effect and the detection signal.

[0054] Figure 9 This is a diagram showing the detection results of a back-side crack using a traditional narrow-pitch probe.

[0055] Figure 10 Comparison of LTR probe detection results for back-side cracks after optimizing the spacing;

[0056] Figure 11 Comparison of detection signals from the optimized LTR probe before EMD reconstruction;

[0057] Figure 12 Comparison of detection signals from the optimized LTR probe after EMD reconstruction. Detailed Implementation

[0058] The present invention will now be described in detail with reference to the embodiments shown in the accompanying drawings:

[0059] Increasing the coil spacing S of an LTR probe can improve deep-layer response. This approach helps reduce the direct coupling between the transmitting coil T and the receiving coil R, making it easier for the receiving coil to capture disturbances from back-side defects. However, its limitations are also significant: the received signal itself is already deeply attenuated, and increasing the spacing significantly enhances the probe's sensitivity to lift-off changes, introducing strong low-frequency lift-off noise. Therefore, simply increasing the spacing often fails to fundamentally improve the signal-to-noise ratio and may even reduce detection stability.

[0060] Empirical Mode Decomposition (EMD), as an adaptive time-frequency analysis method, does not require preset basis functions and is particularly suitable for processing nonlinear and non-stationary detection signals of composite materials, offering advantages over fixed basis function methods such as wavelet transform. Existing research has not specifically addressed the specific noise introduced by hardware parameters (such as increased spacing) to achieve eddy current detection of defects on the back side of composite materials.

[0061] A composite material back defect detection device, comprising: a probe, a signal acquisition module, and a signal processing module.

[0062] like Figure 4a , 4b As shown, the probe is an LTR probe, which includes a transmitting coil T and several receiving coils R. The transmitting coil T and the receiving coils R are arranged side by side with parallel axes. The distance between the transmitting coil T and the receiving coils R is the coil spacing S. The coil spacing S is set to maximize the effective mutual inductance change of the receiving coil R to the defect area on the back of the specimen, that is, to maximize the voltage change ΔV caused by the defect, where ΔV = U2 - U 2ref The coil spacing S corresponds to the peak point of the ΔV-S relationship curve of the voltage change ΔV, where U2 is the output voltage of the receiving coil; U 2ref This is the reference voltage corresponding to the receiving coil in the defect-free region at the starting point of the scan;

[0063] The signal acquisition module is configured to acquire the eddy current detection signal h(t) output by the receiving coil R, which includes the defect signal and lift-off noise;

[0064] The signal processing module is configured to perform Empirical Mode Decomposition (EMD) on the eddy current detection signal h(t), decomposing the eddy current detection signal h(t) into multiple Intrinsic Mode Functions (IMFs) components and a residual term r. n (t); the residual term r n (t) is identified as low-frequency trend term noise caused by lift-off variation and removed; the retained intrinsic mode function (IMF) components are reconstructed to obtain the defect signal with suppressed lift-off noise.

[0065] When the probe has multiple receiving coils (R), the arrangement of each receiving coil (R) and the transmitting coil (T) is as follows: a linear array arrangement, in which each receiving coil is arranged along a straight line and the distance between it and the transmitting coil increases sequentially; or a ring array arrangement, in which each receiving coil is distributed in one or more concentric circles with the transmitting coil as the center; or a matrix array arrangement, in which the transmitting coil and several receiving coils are distributed in a grid pattern in rows and columns.

[0066] The detection method based on the above-mentioned composite material backside defect detection device includes the following steps:

[0067] A1. Determine the coil spacing of the probe: Based on the ply structure of the test specimen and the expected burial depth d and length L of the back defect, adjust the coil spacing S between the transmitting coil T and the receiving coil R of the probe to maximize the effective mutual inductance change of the receiving coil R to the back defect area of ​​the test specimen. The coil spacing S is determined using numerical simulation analysis, including: within the preset coil spacing range S1 < S < S N N is a positive integer. Electromagnetic simulation models of the probe and composite material specimens under different coil spacing conditions are established respectively. The eddy current density distribution, magnetic field penetration depth and induced voltage of the receiving coil at different depths inside the composite material are analyzed under different coil spacing conditions. The amplitude and phase change trends of the detection signal under different coil spacing conditions are compared. Based on the above comparison results, the coil spacing range that achieves a balance between enhancing the response to back defects and reducing the influence of mutual inductive coupling between coils is determined.

[0068] A2. Signal Acquisition: The specimen is scanned using an LTR probe with optimized coil spacing S to acquire the original eddy current detection signal h(t). The acquired original eddy current detection signal h(t) includes the real part and imaginary part of the voltage output from the receiving coil.

[0069] A3. Adaptive Filtering: Perform Empirical Mode Decomposition (EMD) on the original eddy current detection signal h(t), adaptively decomposing it into several intrinsic mode function (IMF) components and a residual term r. n (t), and based on the coil spacing S parameter, adaptively identify and remove low-frequency lift-off noise trend terms related to coil spacing S in the decomposition results. The process of adaptively identifying and removing low-frequency lift-off noise trend terms includes: taking the last residual component r that satisfies the monotonicity condition obtained from Empirical Mode Decomposition (EMD) n (t) is identified as the main noise component related to the lift-off change; further analysis of the frequency characteristics of its preceding intrinsic mode function (IMF) component is conducted. If its energy is mainly concentrated in the extremely low frequency band determined by the probe scanning speed, it is classified as lift-off related noise and removed.

[0070] A4. Defect Signal Reconstruction and Evaluation: The defect signal is reconstructed using the screened intrinsic mode function (IMF) components, and the detection and evaluation of defects on the back side of the composite material are achieved based on their amplitude, phase, or signal-to-noise ratio (SNR). The SNR is calculated using the following method to quantitatively evaluate the detection effect:

[0071]

[0072] Among them, V d V represents the voltage change of the receiving coil in the defect area. n SNR is the voltage change of the receiving coil in a defect-free region, and is V. d The maximum value and V n The difference between the average and V n The maximum value and V n The ratio of the difference between the average and the mean.

[0073] Example 1: In this example, a back-side crack with a relatively long length and a burial depth of 0.5 mm (BC1) and 0.875 mm (BC2) is used as the detection object. The raw eddy current detection signal h(t) output by the receiving coil is acquired by changing the coil spacing. Specifically, the following steps are included:

[0074] Step 1: Geometric Optimization Principle of LTR Probe Coil Spacing

[0075] like Figure 4a , 4b As shown, the structural forms and magnetic field coupling characteristics of several common eddy current probes are compared and analyzed, including those with the transmitting and receiving coils placed perpendicularly to each other, those placed coaxially, and LTR probes. The analysis results show that the LTR probe structure, with the transmitting and receiving coils arranged non-coaxially and laterally in parallel, is beneficial for reducing the mutual inductive coupling between the coils and enhancing the receiving coil's ability to sense eddy current disturbances inside the specimen. Therefore, this type of probe is selected as the eddy current detection probe in this embodiment.

[0076] In eddy current detection, based on the principle of mutual inductance, the output voltage U2 of the receiving coil can be expressed as:

[0077]

[0078] Where j is the imaginary unit, w is the angular frequency, and M 12 M represents the mutual inductance between the transmitting and receiving coils. 2e Z2 represents the mutual inductance between the receiving coil and the test piece, Z2 = R2 + jwL2 is the impedance of the receiving coil, I1 represents the transmitting coil current, I2 represents the receiving coil current, and I... eR1 represents the eddy current loop current, and R2 represents the equivalent resistance of the receiving coil. Based on the probe characteristics, the receiving coil is in an open-circuit state (i.e., I2 = 0), so the above equation simplifies to:

[0079]

[0080] According to the principle of mutual inductance, by M 12 The contributed voltage component mainly reflects the direct coupling between coils, which is an invalid signal for defect detection and will affect the M signal containing structural and defect information. 2e This signal causes interference. Therefore, the purpose of optimizing the LTR probe coil spacing is to enhance M... 2e Decrease M 12 By increasing the coil spacing S, the probe's sensitivity to magnetic field disturbances caused by defects on the back or far surface is enhanced, while direct coupling between the TR coils is suppressed. The amplitude of the receiving coil voltage is obtained as the eddy current detection signal h(t), which contains the defect signal and lift-off noise.

[0081] like Figure 5 , Figure 6a , Figure 6b , Figure 6c The figure shows the spatial magnetic field distribution of an absolute value type probe and LTR probes with different coil spacings. Figure 5 and Figure 6a , Figure 6b , Figure 6c This indicates that for absolute value probes, the receiving coil has the most magnetic field lines passing through it, and the coupling between the surface coils is the strongest, while the mutual inductance between the coils of LTR probes is relatively weak. However, Figure 6a The image shows a traditional narrow-pitch LTR probe, compared to... Figure 6b and Figure 6c Increasing the coil spacing of the probe Figure 6a The mutual inductance between the coils of traditional narrow-pitch LTR probes remains relatively strong. Observations have shown that by increasing the coil spacing of traditional narrow-pitch LTR probes, the magnetic flux passing through their receiving coils can be reduced, thereby reducing the mutual inductance coupling between the coils. This allows the receiving coils to more effectively detect magnetic field disturbances caused by defects on the far surface or back side.

[0082] Therefore, increasing the coil spacing within a certain range can effectively suppress coupling interference between TR coils without sacrificing defect detection sensitivity. However, increasing the coil spacing inevitably enhances the probe's sensitivity to lift-off changes, introduces low-frequency lift-off noise, and reduces the signal amplitude of the receiving coil; therefore, an optimal spacing range needs to be determined. Simultaneously, the eddy current distribution disturbances within the specimen caused by different crack lengths L and burial depths d of far-surface defects exhibit significant differences, thereby altering the local magnetic field and the induced voltage of the receiving coil. Therefore, the selection of the optimal spacing must consider both crack length and burial depth.

[0083] Therefore, by adjusting the spacing S between the transmitting and receiving coils, the influence of mutual inductive coupling between the coils can be reduced to a certain extent, while enhancing the response of the receiving coil to magnetic field disturbances caused by defects inside the composite material, especially in the back region. Based on the above analysis, this embodiment obtains the output signal h(t) of the receiving coil by changing the coil spacing S. This signal simultaneously contains defect response information and low-frequency disturbance components introduced by the lift-off change.

[0084] Step 2: Analysis of coil spacing parameters based on numerical simulation

[0085] Based on the above geometric optimization principle analysis, this embodiment uses numerical simulation to analyze the detection characteristics under different coil spacing conditions. Taking a carbon fiber reinforced polymer (CFRP) laminate as an example specimen, various back crack sizes and burial depth parameters are set, and while keeping the probe lift-off and excitation conditions consistent, the influence of coil spacing variation on magnetic field distribution and detection signal is studied.

[0086] The laminate was constructed from eight layers of unidirectional prepreg, each layer being 0.125 mm thick. The specimen size was 80 mm × 80 mm, and the layup sequence was [0 / 90]4. The anisotropic conductivity parameters were set as follows: 29940 S / m along the fiber direction (x-axis), 4.8 S / m perpendicular to the fiber direction (y-axis), and 1.1 S / m along the thickness direction (z-axis).

[0087] Various sizes of crack defects were pre-designed on the back of the laminate to analyze the impact of defect size and burial depth on detection performance. Specific defect parameters included: BC (10 × 0.4 × 0.25 mm). 3 ), BC1 (10×0.4×0.5 mm) 3 ), BC2 (10×0.4×0.125 mm) 3 ), BC3 (5×0.4×0.5 mm) 3 ) and BC4 (5×0.4×0.125 mm) 3 ).

[0088] The LTR probe was placed at the center of the laminate, with a lift-off distance of 0.5 mm between the probe and the specimen surface. The transmitting and receiving coils had the same structural parameters: an inner diameter of 1.2 mm, an outer diameter of 3.2 mm, a height of 0.8 mm, and 140 turns each. The coil spacing S varied from 0 to 10 mm in 2 mm increments.

[0089] In this invention, a relatively long back-side crack with a burial depth of 0.5 mm (BC1) and 0.875 mm (BC2) is used as the detection object. The raw eddy current detection signal h(t) output by the receiving coil is acquired by changing the coil spacing. During the detection process, the LTR probe scans along the x-axis in 1 mm steps within a range of ±15 mm. Using the signal from the undamaged area as a reference, the effective mutual inductance change between the receiving coil and the back-side defect area is calculated and plotted, i.e., the voltage change value ΔV caused by the defect.

[0090] Simulation results show that the amplitude of the detection signal output by the receiving coil reaches its maximum when the coil spacing S is 6 mm. When the coil spacing is less than 6 mm, the strong mutual inductance coupling between the transmitting and receiving coils dominates, and the magnetic field disturbance caused by the defect is significantly weakened. When the coil spacing increases to 8–10 mm, the attenuation effect during the propagation of the magnetic field inside the conductor is enhanced, the receiving coil's ability to sense the defect signal decreases, and the amplitude of the detection signal decreases accordingly. Therefore, both excessively small and excessively large coil spacings reduce the detection sensitivity of the LTR probe to the magnetic field disturbance of the back defect.

[0091] Further analysis revealed that when the coil spacing was small, the detection signal of the elongated crack was mainly dominated by the tip effect at both ends of the crack, and the signal distribution exhibited a local "double-peak" characteristic centered on the two ends of the crack. By appropriately increasing the coil spacing, the influence of the crack tip effect on the detection signal could be effectively suppressed, thereby enhancing the imaging effect of the crack.

[0092] Under the same detection conditions, back-side cracks with shorter lengths and burial depths of 0.5 mm (BC3) and 0.875 mm (BC4) were scanned. Simulation results show that for short cracks, due to the superposition of magnetic field disturbances generated at both ends of the crack in the crack center region, the optimal coil spacing is relatively small, and the small-pitch LTR probe exhibits higher detection sensitivity for this type of defect. Meanwhile, within a certain burial depth range, the optimal coil spacing increases synchronously with the increase of crack burial depth.

[0093] In summary, the optimal coil spacing of the LTR probe is mainly determined by the crack length and burial depth, and is not very sensitive to changes in probe lift-off. Based on this principle, by arranging multiple receiving coils along the scanning direction to form an array-type eddy current probe, parallel detection can be achieved under different effective coil spacing conditions. This improves the detection capability of back-side cracks of different sizes and burial depths in composite materials without prior knowledge of the defect size.

[0094] Furthermore, by analyzing the amplitude variation of the output signal of the receiving coil, the coil spacing corresponding to the optimal state of mutual inductive coupling between the receiving coil and the conductor under test can be determined. As the coil spacing increases, the receiving coil can sense eddy current signals that penetrate deeper into the composite material and have a longer lateral diffusion path, thereby introducing a phase delay characteristic determined by the coil spacing.

[0095] In summary, comparative analysis of scanning signals under different crack lengths and burial depths reveals a correlation between the coil spacing and the defect geometry and burial depth within a certain parameter range. Based on this, a suitable coil spacing range can be selected for a specific detection object through numerical analysis or experimental calibration, thus balancing detection sensitivity and signal stability.

[0096] In optimizing the coil spacing, the focus is on the combined effects of varying coil spacing on the distribution of the excitation magnetic field, the attenuation characteristics of eddy current density along the thickness direction, and the characteristics of the received coil's induced signal. By comparing the voltage change ΔV caused by defects under different coil spacing conditions, it was found that when the coil spacing is too small, mutual inductive coupling between the coils dominates, and the defect signal is easily submerged; while when the coil spacing is too large, although the magnetic field penetration capability is enhanced, the amplitude of the received signal is significantly attenuated. Therefore, numerical simulations were used to determine the coil spacing range that achieves a balance between reducing mutual inductive coupling and maintaining sufficient defect response.

[0097] Step 3: Lift-off Noise Suppression Method Based on EMD Mode Reconstruction

[0098] Based on geometrically optimized coil spacing, and considering the propagation characteristics of electromagnetic fields, increasing the coil spacing inevitably weakens the sensing signal of the receiving coil. It also inevitably makes the LTR probe more sensitive to lift-off changes, further increasing the impact on weak defect signals. When the probe tilts during testing due to unavoidable factors such as mechanical vibration or uneven specimen surface, at the same tilt angle... Below, increase the coil spacing (S1) S2), which makes the lift-off distance of the wide-pitch LTR probe significantly greater than that of the narrow-pitch LTR probe, i.e., l2>l1, such as Figure 7 As shown. Due to the special structural characteristics of composite materials, the electrical properties of CFRPs with different layups differ significantly, resulting in significant differences in the detection signals on each layup under the same lift-off effect. For example, with a lift-off of 1 mm, the detection voltage on the [0 / 90]4 layup is approximately 35 mV, which is much larger than the detection voltage on the [0]8 layup, which is approximately 0 mV. Figure 8 As shown. Therefore, for composite materials with a special structure like CFRP, the effect of pull-out variation on the detection signal on different ply plates varies significantly, increasing the uncertainty of back-side defect detection.

[0099] While alternative methods such as wavelet filtering can be used, these methods rely on pre-selected basis functions and are difficult to apply to the signal response of highly nonlinear, non-stationary CFRP composites that are closely related to the ply structure. In particular, wavelet methods require the selection of a mother wavelet and a corresponding thresholding strategy, and for CFRP laminates with different plies, there is currently no universal wavelet basis that can reliably characterize their eddy current response signals.

[0100] Therefore, this embodiment introduces the Empirical Mode Decomposition (EMD) method to adaptively process the detection signal. The original eddy current detection signal h(t) output by the receiving coil is decomposed into several intrinsic mode functions (IMFs) and a residual term r. n (t). Among them, the residual term and some low-frequency intrinsic mode function (IMF) components mainly reflect the low-frequency trend characteristics caused by the lift-off variation.

[0101] The algorithm flow based on Empirical Mode Decomposition (EMD) includes the following:

[0102] a) Obtain the original eddy current detection signal h(t)=x(t);

[0103] b) Find all extreme points in x(t) and construct the upper and lower envelopes, and calculate the envelope mean m(t);

[0104] c) Calculate h k (t) = h(t) - m(t);

[0105] d) If h k If (t) satisfies the intrinsic mode function (IMFs) condition, then it is denoted as IMF. n (t), otherwise h k (t) is the new signal. Repeat steps (b)-(c);

[0106] e) Subtract the extracted intrinsic mode functions (IMFs) to obtain the residual r. n (t):

[0107]

[0108] f) If the residual r n If h(t) is not a monotonic function, then treat it as a new h(t) and repeat steps (b)-(e).

[0109] g) Identify low-frequency noise and separate it to the residual trend term r n In (t), removal effectively suppresses lift-off noise while preserving defect signal characteristics:

[0110]

[0111] In this embodiment, residual terms that satisfy the monotonicity condition are identified as noise related to the main trend term of lift-off and removed during the reconstruction process. Subsequently, the retained intrinsic mode functions (IMFs) are reconstructed to obtain the defect response signal after suppressing the lift-off noise. This processing method can adaptively separate low-frequency disturbances and defect features without relying on preset filtering parameters, which is beneficial to improving the stability and signal-to-noise ratio of the detection signal.

[0112] Furthermore, the above embodiments can also be extended to perform frequency characteristic analysis on the intrinsic mode function (IMF) component immediately preceding the residual term to determine whether its energy distribution is mainly concentrated in the extremely low frequency range determined by the probe scanning speed; when the spectral characteristics of the IMF component are dominated by quasi-DC or extremely low frequency components, and its changing trend is consistent with the lift-off change caused by probe scanning, the IMF component is identified as a lift-off related noise component and removed during signal reconstruction.

[0113] Practice shows that the eddy current disturbance caused by back cracks usually manifests as a sudden change in local amplitude or phase during the scanning process. Its frequency characteristics are significantly higher than the quasi-DC trend term caused by lift-off changes. Therefore, by combining the above frequency characteristics and trend analysis, misjudgment of the real defect signal can be effectively avoided.

[0114] Step 4: Experimental Verification

[0115] To verify the effectiveness of the method, this embodiment built an eddy current testing experimental platform to test CFRP laminate specimens containing back cracks.

[0116] During the testing process, the LTR probe is placed above the test piece, and the motion controller drives the displacement platform to move the probe to the non-destructive area of ​​the test piece. Using the automatic bias adjustment function of the lock-in amplifier, the voltage signal output from the receiving coil in this non-destructive area is zeroed to eliminate the influence of the initial bias on the test results. One output of the signal generator is connected to the transmitting coil as the excitation signal source, and the other is used as the reference signal for the lock-in amplifier. The host computer sends control commands to the motion controller via LabVIEW software to control the movement of the displacement platform, thereby driving the LTR probe to scan and test the designated area of ​​the test piece according to the preset scanning path. During the testing process, the lock-in amplifier is used to extract the weak eddy current detection signal and output the real part Vx and imaginary part Vy of the voltage signal. The output voltage amplitude ΔV is calculated based on the real and imaginary parts of the signal, and then transmitted to the host computer for storage and processing via a data acquisition card.

[0117] In the experimental verification of this invention, two CFRP laminate specimens were used as the test objects. Both specimens were prepared using sixteen layers of USN12500 unidirectional prepreg with a resin content of 33%, and the layup sequence was [45 / 0 / -45 / 90]. 2S The specimens measured 250 mm × 200 mm × 2 mm. Multiple narrow cracks, each 0.5 mm wide, were artificially machined on the back of both specimens. Specimen 1 contained two cracks, C2 and C3, with lengths of 23 mm and 22 mm respectively, and embedment depths of 1.0 mm and 1.5 mm respectively. Specimen 2 contained two cracks, C4 and C5, with lengths of 17 mm and 14 mm respectively, and embedment depths of 1.0 mm and 1.5 mm respectively.

[0118] First, C-scans were performed on cracks C2 and C3 in specimen 1 using a conventional narrow-coil-pitch (S=1.8 mm) LTR probe. The corresponding C-scan images are shown below. Figure 9 As shown in the figure, the contrast between the crack area and the background area is low, and the crack features are not obvious. To further verify the detection performance of the method proposed in this invention, an LTR probe with optimized coil spacing (S=6 mm) was used to detect cracks C2 and C3 in specimen 1. The C-scan results are shown in the figure. Figure 10 As shown in the figure. The results show that the geometry-optimized LTR probe can significantly enhance its ability to sense magnetic field disturbances caused by back cracks, and can also achieve stable and reliable detection of back cracks with a burial depth of more than 1 mm.

[0119] Subsequently, an LTR probe with optimized coil spacing (S=6 mm) was used to detect cracks C4 and C5 in specimen 2, and the corresponding C-scan images are shown below. Figure 11 As shown. Observations revealed that increasing the coil spacing improved the sensitivity of back-side defect detection while simultaneously enhancing the probe's sensitivity to changes in lift-off. Figure 11 The detection results show obvious amplitude gradient distributions along the positive x and y axes, which are more pronounced on the right side of the scanned area. This type of low-frequency noise may mask the signal characteristics of defects that are small in size or deep in burial.

[0120] To address the aforementioned problems, this invention combines a geometrically optimized LTR probe with the Empirical Mode Decomposition (EMD) modal reconstruction method. Figure 11 The original eddy current detection signal is adaptively filtered. By performing empirical mode decomposition on the eddy current detection signal and reconstructing the intrinsic mode components related to the defect, the detection result after suppressing noise lift-off is obtained, such as... Figure 12 As shown. The signal-to-noise ratio (SNR) of the detected signals before and after reconstruction is further calculated to quantitatively evaluate the final detection performance. The SNR is obtained using the following formula:

[0121]

[0122] In this embodiment Figure 11 and Figure 12 Based on the signal, the signal-to-noise ratios of cracks C4 and C5 before EMD reconstruction, calculated according to the above formula, are 1.65 ( Figure 11 The upper dashed box) and 0.73 ( Figure 11 The lower half of the dashed box), the reconstructed signal-to-noise ratio is 2.77 ( Figure 12 The upper dashed box) and 1.65 ( Figure 12 (The upper part of the dashed box).

[0123] In summary, this embodiment achieves stable detection of back-side defects in composite materials by geometrically optimizing the LTR probe coil spacing and combining it with an adaptive signal processing method based on EMD modal reconstruction. This technical solution enhances the response capability to back-side defects while effectively suppressing low-frequency lift-off noise introduced by variations in coil spacing, demonstrating good adaptability and practicality.

[0124] The above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made in accordance with the spirit of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A detection method based on a composite material backside defect detection device, the composite material backside defect detection device comprising: The probe includes a transmitting coil and several receiving coils. The transmitting coil and several receiving coils are in the same plane and their axes are parallel. The distance between the transmitting coil and the receiving coil is called the coil spacing, denoted as S. The coil spacing is set to maximize the effective mutual inductance change caused by the receiving coil to the defect area on the back of the specimen while suppressing the background field of mutual inductance coupling between the transmitting coil and the receiving coil. A signal acquisition module is configured to acquire a detection signal output by the receiving coil that includes a defect signal and lift-off noise; A signal processing module is configured to perform empirical mode decomposition on the detection signal to identify and suppress low-frequency lift-off noise related to the coil spacing; The detection method is characterized by comprising the following steps: A1. Determine the coil spacing of the probe: Based on the ply structure of the test specimen and the expected depth and length of the back defect, adjust the coil spacing between the transmitting coil and the receiving coil of the probe to maximize the effective mutual inductance change caused by the receiving coil to the back defect area of ​​the test specimen. A2. Signal Acquisition: The specimen is scanned using a probe with optimized coil spacing to acquire the raw eddy current detection signal; A3. Adaptive filtering: Perform empirical mode decomposition on the original eddy current detection signal, adaptively decompose it into several intrinsic mode function components and a residual term, and based on the coil spacing parameter, adaptively identify and remove the low-frequency lift-off noise trend term related to the coil spacing in the decomposition result; A4. Defect Signal Reconstruction and Evaluation: Defect signals are reconstructed using the selected intrinsic mode function components, and the detection and evaluation of defects on the back side of composite materials are achieved based on their amplitude, phase, or signal-to-noise ratio.

2. The detection method according to claim 1, characterized in that: The signal processing module is configured to perform empirical mode decomposition on the detection signal, decomposing the detection signal into multiple intrinsic mode function components and a residual term; and to identify the residual term as low-frequency trend noise caused by lift-off variation and remove it. The retained intrinsic mode function components are reconstructed to obtain a defect signal with suppressed lift-off noise.

3. The detection method according to claim 1, characterized in that: The value of the coil spacing is chosen to maximize the voltage change of the receiving coil, and the voltage change is denoted as... That is, the coil spacing corresponds to the voltage change. The peak point of the relationship curve, where the voltage change is the difference in detection voltage amplitude between defective and defect-free conditions.

4. The detection method according to claim 1, characterized in that: When the probe has multiple receiving coils, the arrangement of each receiving coil and the transmitting coil is as follows: a linear array arrangement, in which each receiving coil is arranged along a straight line and the distance between it and the transmitting coil increases sequentially; or a ring array arrangement, in which each receiving coil is distributed in one or more concentric circles with the transmitting coil as the center; or a matrix array arrangement, in which the transmitting coil and several receiving coils are distributed in a grid pattern in rows and columns.

5. The detection method according to claim 1, characterized in that: Step A1 uses numerical simulation analysis to determine the coil spacing, which includes: within a preset coil spacing range S1 < S < S N N is a positive integer. Electromagnetic simulation models of the probe and composite material specimens under different coil spacing conditions are established respectively. The eddy current density distribution, magnetic field penetration depth and induced voltage of the receiving coil at different depths inside the composite material under different coil spacing conditions are analyzed. The amplitude and phase change trends of the detection signal under different coil spacing conditions are compared. Based on the above comparison results, the coil spacing range that achieves a balance between enhancing the response to back defects and reducing the mutual inductance coupling between coils is determined.

6. The method according to claim 1, characterized in that, In step A2, the acquired raw eddy current detection signal includes the real part and imaginary part of the voltage output from the receiving coil.

7. The method according to claim 1, characterized in that, In step A3, the process of adaptively identifying and removing low-frequency noise trend terms includes: The last residual component that satisfies the monotonicity condition obtained from empirical mode decomposition is identified as the main noise component related to lift-off variation. Further analysis of the frequency characteristics of its preceding intrinsic mode function component reveals that if its energy is mainly concentrated in the extremely low frequency band determined by the probe scanning speed, it is classified as lift-off related noise and removed.

8. The method according to claim 7, characterized in that, In step A3, the algorithm flow based on empirical mode decomposition includes: a) Obtain the original eddy current detection signal h(t)=x(t); b) Find all extreme points in x(t) and construct the upper and lower envelopes, and calculate the envelope mean m(t); c) Calculate h k (t) = h(t) - m(t); d) If h k If (t) satisfies the intrinsic mode function condition, then it is denoted as IMF. n (t), otherwise h k (t) is the new signal, repeat steps (b)-(c); e) Subtract the extracted intrinsic mode functions to obtain the residual r. n (t): ； f) If the residual r n If h(t) is not a monotonic function, then treat it as a new h(t) and repeat steps (b)-(e). g) Identify low-frequency noise and separate it to the residual trend term r n In (t), removal effectively suppresses lift-off noise while preserving defect signal characteristics: ， The residual terms that satisfy the monotonicity condition are identified as the main trend terms and noise related to the lift-off, and removed from the reconstruction process. Then, the retained intrinsic mode function components are reconstructed to obtain the defect response signal after suppressing the lift-off noise.

9. The method according to claim 1, characterized in that, In step A4, the signal-to-noise ratio (SNR) is calculated in the following way to quantitatively evaluate the detection effect: ， Among them, V d V represents the voltage change of the receiving coil in the defect area. n This represents the voltage change of the receiving coil in a defect-free region.

Images