A method for on-line detection of resin impregnation state of carbon fiber composite material

Abstract

The present application relates to carbon fiber composite material forming detection technical field, specifically point to a kind of carbon fiber composite material resin impregnation state online detection method, this method will array eddy current probe be embedded in mould, using time-sharing excitation mode acquisition each probe's eddy current signal, after pre-processing to signal, extract peak voltage, rise time, attenuation coefficient, center frequency, phase difference and quality factor and other characteristic parameters, and input trained BP neural network realizes resin infiltration state identification.Further based on the spatial interpolation reconstruction of resin front based on multiple probe recognition results, generate impregnation front distribution nephogram, and determine dry spot defect in combination with impregnation degree variation trend.The method has the advantages of strong real-time, good anti-interference ability, high recognition accuracy and suitable for online monitoring etc..

Description

Technical Field

[0001] This invention relates to the field of carbon fiber composite material molding and testing technology, specifically to an online detection method for the resin impregnation state of carbon fiber composite materials. Background Technology

[0002] Carbon fiber composites, with their superior properties such as high strength and low density, are widely used in high-end fields such as aerospace and automotive industries. In the RTM (resin transfer molding) or autoclave processes for preparing carbon fiber composites, the resin impregnation of carbon fibers occurs in a closed environment, where the internal state is highly unpredictable. This closed and unpredictable nature makes precise control of the manufacturing process difficult, leading to significant process redundancy, resource waste, and a high susceptibility to manufacturing defects, severely impacting the quality and performance of carbon fiber composites.

[0003] Currently, there is a lack of efficient online detection methods for assessing the resin impregnation state during the manufacturing process of carbon fiber composites. Traditional detection methods, such as ultrasonic testing and X-ray testing, suffer from low efficiency and difficulty in achieving real-time online monitoring, failing to meet the demands of modern industry for efficient and high-quality production of carbon fiber composites. Therefore, developing an online detection method capable of accurately and in real-time detecting the resin impregnation state inside the mold cavity of carbon fiber composites is of great significance for improving the manufacturing process and product quality of carbon fiber composites. Summary of the Invention

[0004] This invention discloses an online detection method for the resin impregnation state of carbon fiber composite materials, which can realize quantitative, real-time and comprehensive detection of the resin impregnation state of carbon fiber composite materials inside the mold cavity, accurately determine the position of the resin impregnation front and its distribution uniformity, and promptly detect manufacturing defects such as dry spots, thereby eliminating process redundancy, optimizing manufacturing parameters, and improving the quality and production efficiency of composite products.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An online detection method for the resin impregnation state of carbon fiber composite materials includes the following steps: Step 1): Embed multiple eddy current probes in an array into the molding die of carbon fiber composite material. Each eddy current probe includes an excitation coil and an induction coil. Number each eddy current probe and record its spatial coordinates in the die coordinate system. Step 2): Connect the excitation coil of each eddy current probe to the excitation end using the multiplexer at the excitation end, and connect the induction coil of each eddy current probe to the multiplexer at the receiving end. Use a time-division excitation method to excite each eddy current probe in the array eddy current probe in sequence, ensuring that only one eddy current probe is in the excitation state at any given time, and collect the induced voltage signal output by the induction coil of the eddy current probe in the excitation state. Step 3): Calculate the resin wetting at the location of each eddy current probe; Step 4): Perform time-domain and frequency-domain feature analysis on the acquired induced voltage signal, and extract time-domain feature parameters and frequency-domain feature parameters. The extracted time-domain feature parameters include peak voltage, rise time, and attenuation coefficient. The extracted frequency-domain feature parameters include resonant frequency, phase difference, and quality factor. Step 5): Input the extracted six feature parameters into the trained BP neural network model, use the BP neural network model to identify the resin impregnation state of the current area, and output the resin impregnation state level at the location of the corresponding eddy current probe. Step 6): Based on the probability values ​​corresponding to the output resin impregnation state levels, select the resin impregnation state with the highest probability value as the final recognition result, perform spatial interpolation reconstruction on the resin impregnation front, and generate a resin impregnation front cloud map for display.

[0006] Furthermore, in step 3), the resin wettability η ij Calculate using the following formula: Log(σ ij )=(1-η ij )×log(σ fiber )+ η ij ×log(σ resin ) In the above formula, σ ij σ is the equivalent conductivity of the carbon fiber composite material at the location of the (i,j)th eddy current probe; fiber The equivalent conductivity when completely unwetted; σ resin η is the equivalent conductivity when fully wetted; ij The value of η ranges from 0 to 1, where η=0 indicates no wetting and η=1 indicates complete wetting.

[0007] Further, in step 4), the signal-to-noise ratio (SNR) is calculated based on the extracted feature parameters, SNR = 20lg(U s / U n ), where U s U is the signal amplitude. n This represents the noise amplitude. When SNR < 20dB, when applying excitation, increase the excitation current I by 10% until SNR ≥ 20dB or I reaches its maximum value of 100mA. Calculate the average wettability ,when When <0.3, adjust the excitation frequency to 100kHz~200kHz; when When the value is ≥0.3, the excitation frequency should be adjusted to 200kHz~800kHz.

[0008] Furthermore, in step 4), time-domain feature parameters are first extracted, and then frequency-domain feature parameters are extracted, wherein the peak voltage U p The maximum value of the induced voltage signal u(t) is collected during the first oscillation period; rise time t r The attenuation coefficient is defined as the time required for the induced voltage signal u(t) to rise from 10% to 90% of its peak value; an exponential fit is performed on the envelope of the induced voltage signal to obtain the attenuation coefficient. a Perform a windowed Fast Fourier Transform on the induced voltage signal to obtain the spectrum U. ij (f), the frequency point with the largest amplitude in the spectrum is the resonant frequency f0; the phase difference φ between the induced signal and the excitation signal is calculated using the cross-correlation method, and the calculation formula is as follows:

[0009] In the above formula, u e ( k ) represents the discrete sequence of excitation signals; τ is the time delay that maximizes the cross-correlation function; According to the formula Calculate the quality factor Q, where Δf is the bandwidth when the spectral amplitude drops to 0.707 times the maximum value.

[0010] Furthermore, in step 5), the resin impregnation state is divided into five levels: when the resin wettability η ≥ 0.95, it is judged as "completely impregnated"; when the resin wettability 0.8 ≤ η < 0.95, it is judged as "few bubbles"; when the resin wettability 0.5 ≤ η < 0.8, it is judged as "many bubbles"; when the resin wettability 0.2 ≤ η < 0.5, it is judged as "partially impregnated"; and when the resin wettability η < 0.2, it is judged as "not impregnated".

[0011] Furthermore, in step 6), the resin impregnation front is obtained as follows: First, all eddy current probe positions are sorted from largest to smallest wettability, and probe points with wettability between 0.4 and 0.6 are selected as candidate front points; then, cubic spline interpolation is performed on the candidate front points to obtain a continuous curve y=f(x); next, the curvature of each point on the continuous curve is calculated, and points with curvature greater than 0.1 mm are removed. -1For the outliers, perform cubic spline interpolation on the remaining points to obtain the final resin impregnation front.

[0012] Furthermore, in step 2), the preset timing sequence of the time-division excitation is to excite each eddy current probe sequentially in row priority order.

[0013] Furthermore, the online detection method also includes a resin impregnation state visualization step, namely: mapping the resin wetting values ​​at all eddy current probe locations to RGB colors, letting η=0 correspond to blue, η=1 correspond to red, and for wetting values ​​between 0 and 1, using bilinear interpolation to generate a continuous color transition between eddy current probe points, thereby obtaining a resin impregnation state cloud map. The resin impregnation state cloud map marks the position of the resin impregnation front, the advancement speed of the resin impregnation state, and the wetting uniformity index.

[0014] Furthermore, the formula for calculating the advancing speed v of the resin impregnation front is: v = ΔL / Δt', where ΔL is the advancing distance of the resin impregnation front within the time Δt'; the formula for calculating the wetting uniformity index γ is: , σ n The standard deviation of the wettability at all eddy current probe locations. This represents the average wettability at all eddy current probe locations.

[0015] Furthermore, when the wettability at any eddy current probe position changes by less than 0.05 in three consecutive scanning cycles, and the position is located after the resin impregnation front, it is determined that there is a "dry spot" defect at that position, and the defect position is marked with a color on the resin impregnation state cloud map to alert the user.

[0016] This invention enables quantitative, real-time, and comprehensive detection of the resin impregnation state of carbon fiber composite materials inside the molding cavity, accurately determining the position and uniformity of the resin impregnation front, and timely detecting manufacturing defects such as dry spots, thereby eliminating process redundancy, optimizing manufacturing parameters, and improving the quality and production efficiency of carbon fiber composite products. Attached Figure Description

[0017] Figure 1 This is a flowchart of the online detection method for the resin impregnation state of carbon fiber composite materials in the embodiments; Figure 2 This is a cross-sectional view of the eddy current probe embedded in the molding die in the embodiment; Figure 3 This is a plan view of the eddy current probe embedded in the molding die in the embodiment; Figure 4 This is a circuit schematic diagram of the microcontroller's time-division excitation and acquisition of sensing signals in the embodiment. Figure 5This is a schematic diagram illustrating the use of a neural network to sense signals for identification in an embodiment. Figure 6 The resin impregnation state and its leading edge contour plot are shown in the examples; Figure 7 This is a schematic diagram of the scanning sequence of the array eddy current probe in the embodiment; Figure 8 This is a voltage contour map of the induction coils in different immersion areas in the embodiment; Figure 9 This is a resin impregnation state cloud map output in the example.

[0018] Figure label: 01. Upper forming mold; 02. Lower forming mold; 03. Carbon fiber composite material; 11. Excitation coil; 12. Induction coil. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0020] This embodiment discloses an online detection method for the resin impregnation state of carbon fiber composite materials, which is based on an array-type eddy current detection method. The overall process is as follows: Figure 1 As shown, the specific steps are explained below: Step 1): Arrange an array of eddy current probes in the carbon fiber composite molding die; Specifically, such as Figure 2 and Figure 3 As shown, each eddy current probe includes an excitation coil 11 and an induction coil 12. Both the excitation coil 11 and the induction coil 12 are spiral coils formed by winding 0.2mm diameter enameled wire. The coil height is 3mm, the inner diameter is 2mm, the outer diameter is 6mm, the inductance is approximately 5.2μH, and the DC resistance is approximately 1.8Ω. Multiple eddy current probes are arranged in an array of m rows and n columns (the attached figure of this embodiment shows 30 probes as an example). The molding die of the carbon fiber composite material 03 is divided into an upper molding die 01 and a lower molding die 02. Each eddy current probe is embedded 0.5mm below the surface of the lower molding die 02, and a high-temperature resistant epoxy resin insulating layer (approximately 0.3mm thick) is filled between the probe and the die surface to ensure that the eddy current probe does not directly contact the carbon fiber prepreg.

[0021] Each eddy current probe was numbered, and its spatial coordinates (x, y, y) in the mold coordinate system were recorded. ij ,y ij), where i is the row number (1≤i≤m) and j is the column number (1≤j≤n). The spatial coordinates of the eddy current probe are determined based on a predefined coordinate system, which takes the lower left corner of the forming mold as its origin and assigns it the initial coordinates (1,1), representing the eddy current probe in the first row and first column. For example... Figure 3 As shown, along the x-axis and y-axis, for each additional row or column of eddy current probes, the row and column index values ​​increase by Δx and Δy respectively. Therefore, the coordinates of any set of eddy current probes can be represented as (1+Δx, 1+Δy), where Δx and Δy are the offsets of the row and column of the probe relative to the starting point, respectively.

[0022] Pre-calibrate the constant K for each eddy current probe ij The calibration method involves measuring the induced voltage on a standard sample with known conductivity. Specifically, the eddy current probe is placed on the surface of a standard sample with known conductivity σ0, an excitation signal of frequency f0 and current I0 is applied, and the effective value of the induced voltage U0 is measured. The constant K... ij The calculation formula is For different eddy current probes, the constant K ij There may be differences due to manufacturing tolerances, and they need to be calibrated separately.

[0023] Step 2): As Figure 4 As shown, a digital control system is established for time-sharing excitation and signal acquisition; Specifically, the STM32 microcontroller is used as the core of the digital control system, and the signal generator is used as the source of the excitation signal. The output of the signal generator is connected to the power amplifier, and the output of the power amplifier is connected to the multiplexer (model MUX36S18) at the excitation end. The multiplexer at the excitation end is connected to the excitation coil 11 of each eddy current probe, and the multiplexer at the signal receiving end is connected to the induction coil 12 of each eddy current probe. The signal acquisition card acquires the induction signal at the receiving end and feeds it back to the host computer. The microcontroller is used to control the switching of the two multiplexer channels.

[0024] The digital control system outputs a sinusoidal signal with a frequency of 100kHz to 800kHz and an amplitude of 10mA to 100mA. In this embodiment, the initial frequency of the signal generator is set to 500kHz, and the initial voltage amplitude is set to 130mV. The multiplexer switches channels according to a preset timing sequence to achieve time-division excitation. At any given time, only one eddy current probe's excitation coil 11 is enabled, while other eddy current probes are in a high-impedance waiting state. The signal emitted by the signal generator is amplified by a power amplifier and switched by the multiplexer at the excitation end to the currently selected excitation coil 11. Then, the microcontroller control system controls the multiplexer at the excitation end to select a specific channel and transmit the signal to the corresponding excitation coil 11. The excitation coil 11 generates an alternating magnetic field on one side of the object being measured. The STM32 generates a timing interrupt through timer TIM2. The time-division switching time interval Δt is set according to the excitation signal frequency, satisfying Δt ≥ 5T, where T is the period of the excitation signal. In this embodiment, when the excitation frequency is 500kHz, the period T = 10μs, and Δt = 50μs is taken. Time-division excitation uses row priority order, such as... Figure 7 As shown, the excitation is sequentially applied to (1,1)→(1,2)→...→(1,5)→(2,1)→...→(6,5). Each channel is held for 50μs before switching to the next channel. The time to complete one full array scan is T_scan=30×50μs=1.5ms, which meets the real-time requirements of online detection. While exciting the (i,j)th eddy current probe, the induced voltage signal u(t) output by the induction coil 12 of that probe is acquired, with a sampling frequency of not less than 10 times the excitation frequency, and no less than 1024 sampling points are acquired for each channel.

[0025] After multiple induction coils 12 on the other side of the object being tested induce a magnetic field, their outputs are connected to the multiplexer at the receiving end. Similarly, the microcontroller control system synchronously controls the selection of the corresponding receiving channel. The selected receiving signal is digitally processed by the signal acquisition card and then transmitted to the host computer for display, storage and analysis. The microcontroller simultaneously controls two multiplexers to ensure that the same channel of the multiplexer is turned on and off at the same time, realizing the synchronous switching of multi-channel excitation and reception, and completing the detection of the object being tested.

[0026] Step 3): Calculate the resin wetting at the location of each eddy current probe; Specifically, the effective value of the acquired induced voltage signal u(t) is calculated to obtain the induced voltage amplitude U. ij Based on the principle of eddy current detection, the equivalent conductivity σ of the carbon fiber composite material at the position of probe (i,j) can be calculated. ij Based on the carbon fiber / resin two-phase hybrid conductivity model, the resin wetting degree η at the (i,j)th probe position can be calculated. ij : Log(σij )=(1-η ij )×log(σ fiber )+ η ij ×log(σ resin ) In the above formula, σ fiber The equivalent conductivity when the fiber is completely unwetted (dry); σ resin σ is the equivalent conductivity when the resin is fully impregnated (completely filled). fiber and σ resin All of these are pre-calibrated constants. η ij The value of η ranges from 0 to 1, where η=0 indicates no wetting and η=1 indicates complete wetting.

[0027] Step 4): Perform time-domain and frequency-domain feature analysis on the acquired induced voltage signal u(t). First, extract the time-domain feature parameters, and then extract the frequency-domain feature parameters. The extracted time-domain feature parameters include the peak voltage U. p Rise time t r attenuation coefficient a The extracted frequency domain feature parameters include resonant frequency f0, phase difference φ, and quality factor Q.

[0028] Specifically, peak voltage U p The maximum value of the induced voltage signal u(t) is collected during the first oscillation period; rise time t r The attenuation coefficient is defined as the time required for the induced voltage signal u(t) to rise from 10% to 90% of its peak value; an exponential fit is performed on the envelope of the induced voltage signal to obtain the attenuation coefficient. a u(t)=A e (- at Perform a windowed Fast Fourier Transform on the induced voltage signal to obtain the spectrum U. ij (f), the frequency point with the largest amplitude in the spectrum is the resonant frequency f0; the phase difference φ between the induced signal and the excitation signal is calculated using the cross-correlation method, and the calculation formula is as follows:

[0029] In the above formula, u e ( k ) represents the discrete sequence of excitation signals; τ is the time delay that maximizes the cross-correlation function; Using formula Calculate the quality factor Q, where Δf is the bandwidth when the spectral amplitude drops to 0.707 times the maximum value.

[0030] The signal-to-noise ratio (SNR) is calculated based on the extracted feature parameters: SNR = 20lg(U s / U n ), where Us U is the signal amplitude. n This represents the noise amplitude. When SNR < 20dB, increase the excitation current I by 10% when applying excitation until SNR ≥ 20dB or I reaches its maximum value of 100mA. Calculate the average wettability of the entire array. , ;when When <0.3, the excitation frequency is adjusted to 100kHz~200kHz (200kHz in this embodiment) to improve the penetration depth; when When the value is ≥0.3, the excitation frequency is adjusted to 200kHz~800kHz (in this embodiment, it is adjusted to 500kHz) to improve the detection resolution.

[0031] Step 5): Input the extracted six feature parameters into the trained BP neural network model, use the BP neural network model to identify the resin impregnation state of the current area, and output the resin impregnation state level at the location of the corresponding eddy current probe. Specifically, such as Figure 5 As shown, this embodiment uses a three-layer BP neural network model, which has an input layer (6 neurons), a hidden layer (15 neurons), and an output layer (5 neurons). The 6 neurons in the input layer correspond to a 6-dimensional feature vector [U... p , t r , a [f0, φ, Q], the five neurons in the output layer correspond to five resin impregnation state levels, namely: when the resin wetting degree η≥0.95, it is judged as "completely impregnated"; when the resin wetting degree 0.8≤η<0.95, it is judged as "few bubbles"; when the resin wetting degree 0.5≤η<0.8, it is judged as "many bubbles"; when the resin wetting degree 0.2≤η<0.5, it is judged as "partially impregnated"; when the resin wetting degree η<0.2, it is judged as "not impregnated".

[0032] The activation function used is the Sigmoid function, and the training algorithm employs the backpropagation algorithm with momentum. During the pre-training of the BP neural network model, induction signals from different resin impregnation states are collected for machine learning. 1800 sets of induction signals are collected under different process conditions (corresponding to 30 probes × 60 sets of process parameters in this embodiment). The actual impregnation state is observed under a microscope and manually labeled. The training parameters are set as follows: learning rate 0.01, momentum factor 0.9, maximum number of iterations 10000, and training target error 0.01. The extracted 6-dimensional feature vector is input into the trained BP neural network model, which outputs the probability values ​​of five states. The state with the highest probability is taken as the recognition result.

[0033] Step 6): Forming a resin impregnation front contour map; Specifically, firstly, all eddy current probes (30 probes in this embodiment) are sorted from largest to smallest wettability, and probe points with wettability between 0.4 and 0.6 are selected as candidate leading edge points. Then, cubic spline interpolation is performed on the candidate leading edge points to obtain the continuous curve y=f(x). Next, the curvature of each point on the continuous curve is calculated, and points with curvature greater than 0.1 mm are removed. -1 For outliers, perform cubic spline interpolation on the remaining points to obtain the final resin impregnation front. Record the position of the resin impregnation front at two adjacent time points, calculate the advancement distance of the front in the resin flow direction, and calculate the advancement velocity v of the resin impregnation front according to the formula v=ΔL / Δt', where ΔL is the advancement distance of the resin impregnation front in time Δt'. The formula for calculating the wetting uniformity index γ is: , σ n The standard deviation of the wettability at all eddy current probe locations. , This represents the average wettability at all eddy current probe locations; the closer γ is to 1, the more uniform the wettability.

[0034] Step 7): Visualization of the resin impregnation state.

[0035] The wetting value at each probe location is mapped to a different RGB color to generate a resin impregnation state cloud map (e.g., Figure 9 As shown in the figure, the resin impregnation state cloud map is displayed in real time on the human-machine interface. The position of the resin impregnation front is marked with a white solid line, and the advancement speed and uniformity index are also displayed. Let η=0 correspond to blue and η=1 correspond to red. For wettability between 0 and 1, a continuous color transition is generated between the eddy current probe points using bilinear interpolation, thus obtaining a 100mm×100mm resin impregnation state cloud map. When the wettability at any eddy current probe position is detected to have a change of less than 0.05 for three consecutive scanning cycles, and the position is located after the resin impregnation front, it is determined that there is a "dry spot" defect at that position. After a "dry spot" is detected, the position is marked with a red flashing mark on the resin impregnation state cloud map to remind the inspection personnel to pay attention.

[0036] Specific molding example: 30 eddy current probes are set, the molding die temperature is kept constant at 60℃, the injection pressure is 0.3MPa, and a carbon fiber pre-molded sports epoxy resin system is used. Figure 7 The sequence of point-by-point excitation and signal acquisition is shown, with a full array scan period of 1.5ms. The continuous monitoring duration covers the entire process from resin injection to complete wetting, and the resulting resin impregnation front state is as follows. Figure 6 As shown, Figure 8 The diagram shows the induced voltage cloud map directly measured by the array probe, and the voltage amplitude distribution at different probe positions is highly consistent with the actual resin impregnation trajectory. Figure 9The image shows the immersion state recognition results of the BP neural network model after the feature vectors of each probe position are input into the model. The image clearly distinguishes the fully immersed area, the area with fewer bubbles, the area with more bubbles, the semi-immersed area, and the unimmersed area. The recognition results are in good agreement with the actual immersion state observed under a microscope.

[0037] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for online detection of the resin impregnation state of carbon fiber composite materials, characterized in that, Includes the following steps: Step 1): Embed multiple eddy current probes in an array into the molding die of carbon fiber composite material. Each eddy current probe includes an excitation coil and an induction coil. Number each eddy current probe and record its spatial coordinates in the die coordinate system. Step 2): Connect the excitation coil of each eddy current probe to the excitation end using the multiplexer at the excitation end, and connect the induction coil of each eddy current probe to the multiplexer at the receiving end. Use a time-division excitation method to excite each eddy current probe in the array eddy current probe in sequence, ensuring that only one eddy current probe is in the excitation state at any given time, and collect the induced voltage signal output by the induction coil of the eddy current probe in the excitation state. Step 3): Calculate the resin wetting at the location of each eddy current probe; In step 3), the resin wettability η ij Calculate using the following formula: Log(σ ij )=(1-n ij )×log(σ fiber )+ the ij ×log(σ resin ) In the above formula, σ ij σ is the equivalent conductivity of the carbon fiber composite material at the location of the (i,j)th eddy current probe; fiber The equivalent conductivity when completely unwetted; σ resin η is the equivalent conductivity when fully wetted; ij The value of η is 0-1, where η=0 indicates no wetting and η=1 indicates complete wetting. Step 4): Perform time-domain and frequency-domain feature analysis on the acquired induced voltage signal, and extract time-domain feature parameters and frequency-domain feature parameters. The extracted time-domain feature parameters include peak voltage, rise time, and attenuation coefficient. The extracted frequency-domain feature parameters include resonant frequency, phase difference, and quality factor. Step 5): Input the extracted six feature parameters into the trained BP neural network model, use the BP neural network model to identify the resin impregnation state of the current area, and output the resin impregnation state level at the location of the corresponding eddy current probe. Step 6): Based on the probability values ​​corresponding to the output resin impregnation state levels, select the resin impregnation state with the highest probability value as the final recognition result, perform spatial interpolation reconstruction on the resin impregnation front, and generate a resin impregnation front cloud map for display. In step 6), the resin impregnation front is obtained as follows: First, all eddy current probe positions are sorted from largest to smallest wettability, and probe points with wettability between 0.4 and 0.6 are selected as candidate front points; then, cubic spline interpolation is performed on the candidate front points to obtain a continuous curve y=f(x); next, the curvature of each point on the continuous curve is calculated, and points with curvature greater than 0.1 mm are removed. -1 For the outliers, perform cubic spline interpolation on the remaining points to obtain the final resin impregnation front.

2. The method for online detection of resin impregnation state of carbon fiber composite materials according to claim 1, characterized in that: In step 4), the signal-to-noise ratio (SNR) is calculated based on the extracted feature parameters, SNR = 20lg(U s / U n ), where U s U is the signal amplitude. n This represents the noise amplitude. When SNR < 20dB, when applying excitation, increase the excitation current I by 10% until SNR ≥ 20dB or I reaches its maximum value of 100mA. Calculate the average wettability ,when When <0.3, adjust the excitation frequency to 100kHz~200kHz; when When the value is ≥0.3, the excitation frequency should be adjusted to 200kHz~800kHz.

3. The method for online detection of resin impregnation state of carbon fiber composite materials according to claim 1, characterized in that: In step 4), time-domain feature parameters are extracted first, and frequency-domain feature parameters are extracted second, wherein the peak voltage U p The maximum value of the induced voltage signal u(t) is collected during the first oscillation period; rise time t r The attenuation coefficient is defined as the time required for the induced voltage signal u(t) to rise from 10% to 90% of its peak value; an exponential fit is performed on the envelope of the induced voltage signal to obtain the attenuation coefficient. a Perform a windowed Fast Fourier Transform on the induced voltage signal to obtain the spectrum U. ij (f), the frequency point with the largest amplitude in the spectrum is the resonant frequency f0; the phase difference φ between the induced signal and the excitation signal is calculated using the cross-correlation method, and the calculation formula is as follows: ； In the above formula, u e ( k ) represents the discrete sequence of excitation signals; τ represents the time delay that maximizes the cross-correlation function; According to the formula Calculate the quality factor Q, where Δf is the bandwidth when the spectral amplitude drops to 0.707 times the maximum value.

4. The method for online detection of resin impregnation state of carbon fiber composite materials according to claim 1, characterized in that: In step 5), the resin impregnation status is divided into five levels. When the resin wettability η ≥ 0.95, it is judged as "completely impregnated"; when the resin wettability 0.8 ≤ η < 0.95, it is judged as "few bubbles"; when the resin wettability 0.5 ≤ η < 0.8, it is judged as "many bubbles"; when the resin wettability 0.2 ≤ η < 0.5, it is judged as "partially impregnated"; and when the resin wettability η < 0.2, it is judged as "not impregnated".

5. The method for online detection of resin impregnation state of carbon fiber composite materials according to claim 1, characterized in that: In step 2), the preset timing sequence of the time-division excitation is to excite each eddy current probe sequentially in row priority order.

6. The method for online detection of resin impregnation state of carbon fiber composite materials according to claim 1, characterized in that: The online detection method also includes a resin impregnation state visualization step, namely: mapping the resin wetting values ​​at all eddy current probe locations to RGB colors, letting η=0 correspond to blue, η=1 correspond to red, and for wetting values ​​between 0 and 1, using bilinear interpolation to generate a continuous color transition between eddy current probe points, thereby obtaining a resin impregnation state cloud map. The resin impregnation state cloud map marks the position of the resin impregnation front, the advancement speed of the resin impregnation state, and the wetting uniformity index.

7. The method for online detection of resin impregnation state of carbon fiber composite materials according to claim 6, characterized in that: The formula for calculating the advancing velocity v of the resin impregnation front is: v = ΔL / Δt', where ΔL is the advancing distance of the resin impregnation front within the time Δt'; the formula for calculating the wetting uniformity index γ is: , σ n The standard deviation of the wettability at all eddy current probe locations. This represents the average wettability at all eddy current probe locations.

8. The method for online detection of resin impregnation state of carbon fiber composite materials according to claim 6, characterized in that: When the wettability at any eddy current probe position changes by less than 0.05 for three consecutive scanning cycles, and the position is located after the resin impregnation front, it is determined that there is a "dry spot" defect at that position, and the defect position is marked with a color on the resin impregnation state cloud map to alert the user.

Images