A method for processing nondestructive testing images to determine impact damage in composite laminates.

Abstract

This invention discloses a processing method for non-destructive testing (NDT) images used to determine impact damage in composite laminates. The method obtains the positions of all pixels satisfying the range conditions by setting the RGB range of pixel colors in the damaged area. Using the impact point in the NDT image as the rotation center, it calculates the actual distance between the edge pixels of the damaged area corresponding to each rotation angle and the impact point according to a predetermined rotation angle step size, thus constructing a rotation angle-radius data sequence. By employing Hampel filtering to identify outliers exceeding the allowable standard deviation range in the rotation angle-radius data sequence, it eliminates split-wire regions with abnormally large radii in the damaged area. This allows for the acquisition of a filtered NDT image based on the filtered rotation angle-radius data sequence, resulting in a more theoretically consistent circular impact damage area and a more reasonable impact damage area result.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing of composite material damage, and more particularly to a method for processing nondestructive testing images to determine impact damage in composite laminates. Background Technology

[0002] Impact damage is a common form of damage to aerospace composite materials, especially composite laminate structures. For structural damage caused by impact, ultrasonic non-destructive testing technology is usually used for flaw detection. Among them, ultrasonic C-scan imaging technology uses the principle of ultrasonic flaw detection to extract the echo signal perpendicular to the direction of the sound beam, forming a cross-sectional image within a certain depth range, which can intuitively show the distribution area of ​​defects.

[0003] In C-scan nondestructive testing images, the pixel colors corresponding to damaged and undamaged areas show a significant difference. By setting the RGB range of the damaged area color using a Red-Green-Blue (RGB) format and counting the number of pixels that fall within this range, the area of ​​the damaged area can be calculated based on the calibrated area of ​​a single pixel. Therefore, the pixel method is commonly used to quantitatively calculate and determine the size and area of ​​the damaged area. For composite laminates, the damaged area is the normal projected area of ​​the damaged area in each layup.

[0004] However, the pixel-based method for calculating the damaged area has limitations. Specifically, when the impact energy is high, the back of the composite laminate, far from the impact surface, is prone to developing positive or negative 45-degree splits. Studies have shown that the impact of back splits on residual strength is negligible; therefore, in the damage tolerance design stage, the damaged area can be simplified to a circular damaged area centered on the impact point. However, compared to the undamaged area, the ultrasonic echo signal at the split point changes, causing the color of the splits to appear the same as the damaged area in the non-destructive testing image, ultimately making it impossible to distinguish between the split area and the circular damaged area. In other words, the damaged area obtained using the pixel-based method includes the area of ​​the back splits. Furthermore, due to the limitation of the ultrasonic probe beam width, the accuracy of the C-scan detection depth is limited, resulting in rough edges of the damaged area in the C-scan non-destructive testing image, which can even lead to distortion of the damaged area.

[0005] Therefore, there is an urgent need for an image processing method that can accurately identify the damaged area and the splitting area in the non-destructive testing image of composite material impact damage, thereby eliminating the splitting area to accurately calculate the area of ​​the damaged area, and at the same time correct the rough edges of the damaged area to remove the influence of high-frequency noise. Summary of the Invention

[0006] Therefore, in order to overcome the problems that existing methods for processing non-destructive testing images of impact damage in composite materials cannot accurately identify the impact damage area, effectively identify and eliminate interference areas such as those caused by wire splitting and noise effects, this invention provides a novel method for processing non-destructive testing images for determining impact damage in composite laminates.

[0007] Specifically, the present invention solves the above-mentioned technical problems through the following technical solutions:

[0008] This invention provides a method for processing non-destructive testing images to determine impact damage in composite laminates, characterized in that the method includes the following steps:

[0009] Set the RGB range of the pixel color in the damaged area, and obtain the positions of all pixels that meet the range conditions;

[0010] Using the impact point in the non-destructive testing image as the rotation center, the actual distance between the edge pixel of the damage area corresponding to each rotation angle and the impact point is calculated according to the predetermined rotation angle step size, and used as the radius corresponding to that rotation angle, thereby constructing a rotation angle-radius data sequence;

[0011] Hampel filtering was used to identify outliers in the rotation angle-radius data sequence that exceeded the allowable standard deviation range, in order to eliminate the splitting areas with abnormal radius sizes in the damaged area;

[0012] Based on the filtered rotation angle-radius data sequence, the filtered non-destructive testing image is obtained.

[0013] The non-destructive testing image processing method disclosed in this invention for determining impact damage in composite laminates converts the non-destructive testing image into a rotation angle-radius data sequence, facilitating direct filtering of this data sequence. Furthermore, it establishes a correspondence between the image and the actual location of the damaged area on the composite laminate, and introduces the Hampel filtering method to filter the obtained data sequence, thereby eliminating the influence of wire splitting on the size of the damaged area displayed in the image, thus obtaining an accurate image of the damaged area and a more reasonable impact damage area result.

[0014] According to one embodiment of the present invention, the impact point is the geometric center point of the damaged region obtained by the positions of all pixels within the damaged region in the non-destructive testing image. By setting the impact point as the initially obtained geometric center point of the damaged region, abnormal edge pixels with abrupt changes in distance can be quickly and effectively identified based on the distance between the edge pixels and the impact point, thereby identifying the split-fiber region.

[0015] According to another embodiment of the present invention, the step of calculating the actual distance between the edge pixel and the impact point corresponding to each rotation angle includes: dimensional calibration of the nondestructive testing image to obtain the actual size corresponding to each pixel; calculating the pixel distance between the edge pixel and the impact point by multiplying the pixel distance by the actual size. By dimensional calibration of the nondestructive testing image, a relationship between the size of the damaged area in the image and the size of the damaged area in the actual physical object can be established, thereby enabling the location and size information of the damaged area in the actual physical object to be obtained from the image.

[0016] According to another embodiment of the present invention, the steps for identifying outliers using the Hampel filtering method include: setting a window range and constructing a window data sequence corresponding to each data point in the rotation angle-radius data sequence; calculating a deviation factor for each data point based on the window data sequence and the rotation angle-radius data sequence; and determining that the data point corresponding to the deviation factor is an outlier if the deviation factor exceeds the allowable standard deviation range. The Hampel filtering method for identifying outliers allows for adjustment of the window range to change the size of the constructed window data sequence, thereby altering the performance requirements for computers and exhibiting strong universality. Furthermore, users can modify the allowable standard deviation range based on experience, making the identification and correction of outliers more closely aligned with reality.

[0017] According to another embodiment of the present invention, the step of identifying outliers using the Hampel filtering method further includes: based on the inverse Gaussian error function, obtaining the deviation factor corresponding to each data point in the rotation angle-radius data sequence according to the absolute value of the difference between each data point in the rotation angle-radius data sequence and the median of the corresponding window data sequence, and the median of the difference data sequence of the window data sequence.

[0018] According to another embodiment of the present invention, the step of identifying outliers using the Hampel filtering method further includes: calculating the absolute value of the difference between each data point in each window data sequence and its median, to obtain a difference data sequence.

[0019] According to another embodiment of the present invention, the step of identifying outliers using the Hampel filtering method further includes: when the data point corresponding to the deviation factor is determined to be an outlier, replacing the data of the outlier with the median of the window data sequence corresponding to the outlier's data, thereby eliminating the split-wire region with an abnormal radius in the damaged area. This method of correcting outliers by replacing the outlier's data with the median of the window data sequence not only eliminates the influence of the split-wire region but also retains reliable data at that rotation angle, thus ensuring that a reasonable damaged area size can be obtained.

[0020] According to another embodiment of the present invention, the processing method further includes: using a fast Fourier transform low-pass filtering method to filter the filtered rotation angle-radius data sequence to eliminate the influence of high-frequency noise.

[0021] According to another embodiment of the present invention, the steps of filtering using the fast Fourier transform low-pass filtering method include: setting a preset sampling frequency to transform the rotation angle-radius data sequence processed by the Hampel filtering method from the time domain space to the frequency domain space; and using a low-pass filter with a predetermined cutoff frequency to perform low-pass filtering on the rotation angle-radius data sequence in the frequency domain, thereby smoothing the edges of the damaged area to eliminate the influence of high-frequency noise.

[0022] According to another embodiment of the present invention, the step of filtering using the fast Fourier transform low-pass filtering method further includes using inverse fast Fourier transform to transform the rotation angle-radius data sequence after eliminating high-frequency noise to the time domain, thereby constructing the filtered rotation angle-radius data sequence.

[0023] According to another embodiment of the present invention, the processing method further includes: calculating all pixels within a set RGB range of the filtered nondestructive testing image, wherein these pixels are the damaged region pixels; and calculating the sum of the sizes of all damaged region pixels to obtain the area of ​​the damaged region. Based on the nondestructive testing image after eliminating the effects of the splitting region and high-frequency noise, the accurate damaged region in the image can be obtained. Then, based on the relationship between the nondestructive testing image and the actual size of the object, the accurate actual area of ​​the damaged region is obtained.

[0024] According to another embodiment of the present invention, the processing method further includes: processing multiple nondestructive testing images to obtain the damage area corresponding to each nondestructive testing image, and fitting a damage area-residual strength curve based on the obtained damage area and its corresponding compressive residual strength data.

[0025] Based on common knowledge in the field, the above-mentioned preferred conditions can be combined arbitrarily to obtain various preferred embodiments of the present invention.

[0026] The beneficial technical effects and advantages that can be achieved by the non-destructive testing image processing method for determining impact damage of composite laminates according to the above embodiments of the present invention are as follows:

[0027] This method for processing non-destructive testing (NDT) images of impact damage in composite laminates converts the NDT images of impact damage into a rotation angle-radius data sequence. This establishes a relationship between the position and size of the image and the actual damaged area of ​​the composite laminate, facilitating direct filtering of the data sequence to optimize the NDT image processing. Furthermore, the Hampel filtering method is introduced to identify outliers in the data sequence, thereby identifying and eliminating the influence of the splitting region on the damaged area and its area.

[0028] Simultaneously, a fast Fourier transform low-pass filtering method is introduced to process the filtered data sequence, thereby optimizing the edge smoothness of the damaged area to reduce the influence of high-frequency noise and ensure the realism of the damaged area image. This overcomes the limitations of current ultrasonic probes in terms of beam width and detection depth accuracy. Furthermore, the non-destructive testing image obtained by this method is a normal projection image of the damaged area of ​​each layup, and the influence of the splitting region and high-frequency noise is eliminated. Therefore, this method is suitable for processing non-destructive testing images of impact damage in composite laminates with various layups, dimensions, and environmental conditions, demonstrating strong universality. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating a method for processing images of non-destructive testing to determine impact damage in composite laminates according to a preferred embodiment of the present invention.

[0030] Figure 2 This is a schematic diagram of the non-destructive testing image of impact damage to a composite laminate according to a preferred embodiment of the present invention, within the RGB range of pixel color in the damaged area.

[0031] Figure 3 for Figure 2 The diagram shown is a schematic of the non-destructive testing image after being filtered using the Hampel filtering method.

[0032] Figure 4 for Figure 3 The diagram shown is a schematic of the non-destructive testing image after being filtered by the Fast Fourier Transform low-pass filtering method.

[0033] Figure 5 This is a fitted curve of the damaged area versus the remaining intensity obtained after processing multiple nondestructive testing images according to a preferred embodiment of the present invention. Detailed Implementation

[0034] The preferred embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. The following description is exemplary and not intended to limit the present invention. Any other similar situations also fall within the protection scope of the present invention.

[0035] In the following detailed description, sequential terms such as "first," "next," "following," "then," "again," etc., are used in the order of steps described with reference to the accompanying drawings. The specific steps of embodiments of the present invention can be arranged in a variety of different orders, and the sequential terms are for illustrative purposes and not for limitation.

[0036] Impact damage is a common form of damage in aerospace composite laminate structures. For structural damage caused by impact, ultrasonic non-destructive testing technology is usually used for flaw detection. Among them, ultrasonic C-scan imaging technology uses the principle of ultrasonic flaw detection to extract the echo signal perpendicular to the direction of the sound beam, forming a cross-sectional image within a certain depth range, which can intuitively show the distribution area of ​​defects.

[0037] In C-scan nondestructive testing images, the pixel colors corresponding to damaged and undamaged areas show a significant difference. By setting the RGB range of the damaged area's color using a Red-Green-Blue (RGB) format, counting the number of pixels that fall within this range, and then calculating the area of ​​the damaged area based on the calibrated area of ​​each pixel, the pixel method is commonly used to quantitatively calculate and determine the size and area of ​​the damaged area. For composite laminates, this damaged area is the normal projected area of ​​the damaged area in each layup.

[0038] The pixel-based method for calculating the damaged area has limitations. When impact energy is high, the back of the composite laminate, far from the impact surface, is prone to developing splitting at ±45°. Studies show that the impact of back splitting has a negligible effect on residual strength; therefore, in the damage tolerance design stage, the impact damage area can be simplified to a circular damage area centered on the impact point. However, compared to the undamaged area, the ultrasonic echo signal at the splitting point changes, causing the splitting to appear as the color of the damaged area in the non-destructive testing image, making it impossible to distinguish between the splitting area and the circular damage area. Therefore, the damaged area area obtained using the pixel-based method will additionally include the area of ​​the back splitting region.

[0039] In addition, due to the limitation of the ultrasonic probe beam width, the accuracy of C-scan detection depth is limited. Therefore, the non-destructive testing images obtained by C-scan are affected by high-frequency noise, which makes the edges of the damaged area obtained by non-destructive testing appear rough. If the edges are rough enough, it may even cause the damaged area to be distorted.

[0040] Therefore, there is a need for an image processing method that can accurately identify the damaged area and the splitting area in the non-destructive testing image of composite material impact damage, thereby eliminating the splitting area to accurately calculate the area of ​​the damaged area, and at the same time correct the rough edges of the damaged area to remove the influence of high-frequency noise.

[0041] To address the aforementioned problems, this invention provides a method for processing non-destructive testing (NDT) images to determine impact damage in composite laminates. This method obtains the positions of all pixels satisfying the range conditions by setting the RGB range of pixel colors in the damaged area. Using the impact point in the NDT image as the rotation center, it calculates the actual distance between the edge pixels of the damaged area and the impact point according to a predetermined rotation angle step size, thus using this distance as the radius corresponding to that rotation angle, thereby constructing a rotation angle-radius data sequence. Subsequently, a Hampel filtering method is used to identify outliers in the rotation angle-radius data sequence that exceed the allowable standard deviation range, eliminating abnormally sized split areas in the damaged area. This allows for the acquisition of a filtered NDT image based on the filtered rotation angle-radius data sequence, thereby obtaining an accurate damaged area. The specific process will be further detailed below with reference to the accompanying drawings.

[0042] Figure 1 A flowchart illustrating a method for processing non-destructive testing images to determine impact damage in composite laminates according to a preferred embodiment of the present invention is shown. First, an original non-destructive testing image for determining impact damage in the composite laminate is obtained using C-scan imaging technology. Subsequently, as... Figure 1 As shown, the RGB range of the pixel colors in the damaged area is set, and the positions of all pixels satisfying the range condition are obtained. Then, a specific rotation angle step size θ is set, and with the impact point in the image as the rotation center, the edge pixels of the damaged area corresponding to each rotation angle are sequentially obtained. The actual distance between the edge pixels and the impact point is calculated as the radius corresponding to the current rotation angle, constructing a rotation angle-radius data sequence R. The control parameter for this process is the rotation angle step size θ. By controlling the rotation angle step size, the data volume of the rotation angle-radius data sequence and the resolution relative to the image sampling can be adjusted.

[0043] The impact point is the geometric center of the damage region obtained by locating all pixels within the damage area in the non-destructive testing (NDT) image. Furthermore, this processing method calibrates the NDT image to obtain the actual size of each pixel, then calculates the pixel distance between edge pixels and the impact point. By multiplying the pixel distance by the actual size, the actual distance between the edge pixels and the impact point corresponding to each rotation angle is obtained. This establishes a relationship between the damage region in the NDT image and the actual location and size of the impact damage in the composite laminate.

[0044] Next, the Hampel filtering method is used to identify outliers in the rotation angle-radius data sequence R that exceed the allowable standard deviation range, in order to eliminate the splitting areas with abnormal radius sizes in the damaged region. Specifically, the rotation angle-radius data sequence R = {r1, r2, ..., r...} nSet the window range in the filtering parameters to k, and apply the filter to any data r in the rotation angle-radius data sequence R. n Construct the corresponding window data sequence X n ={x1, x2, ..., x 2k+1}, then find the window data sequence X n The median, denoted as median(X) n For example, by processing the window data sequence X n Sort the data in the data in ascending order, and find the sorted window data sequence X. n The data in the middle position is the median.

[0045] Based on the obtained window data sequence X n and the window data sequence X n The corresponding median, calculate the window data sequence X. n The absolute value of the difference between each data point (or element) and its median is used to obtain the corresponding difference data sequence, denoted as the difference data sequence Y. n Specifically, the difference data sequence Y is obtained by calculating using the formula shown below. n :

[0046] Y n ={y1, y2, ..., y n}={|x1-median(X n )|,…,|x 2k+1 -median(X n )|} (1)

[0047] Then find the difference data sequence Y n The median, denoted as median(Y) n For example, by analyzing the difference data sequence Y... n Sort the data in the dataset according to size, and find the sorted difference data sequence Y. n The data in the middle position is the median.

[0048] The deviation factor for any data point is calculated based on the window data sequence and the rotation angle-radius data sequence, that is, based on any data point r. n Its corresponding window data sequence X n and the difference data sequence Y n Obtain the corresponding deviation factor H n Specifically, based on the inverse Gaussian error function erf -1 (x), based on any data r in the rotation angle-radius data sequence. n The absolute value of the difference between the median of its corresponding window data sequence and |r n-median(X n The median (Y) of the difference data sequence of the window data sequence. n The rotation angle-radius data sequence r is obtained by calculating using the following formula (2). n Corresponding deviation factor H n :

[0049]

[0050] The allowable standard deviation range for the filter parameters is set to nσ. If the deviation factor exceeds the allowable standard deviation range, the data point corresponding to that deviation factor is determined to be an outlier. That is, when H... n When H > nσ, it is considered that H n The corresponding r n This is considered an outlier. Furthermore, when a data point corresponding to this deviation factor is determined to be an outlier, the outlier's data is replaced with the median of the window data sequence corresponding to that outlier, thus eliminating the split-wire region with an abnormal radius within the damaged area. In other words, when r is determined... n When it is an outlier, use median(X) n ) replace r n This is to eliminate the abnormally sized splitting areas in the damaged region.

[0051] The processing method also includes using a Fast Fourier Transform low-pass filter to filter the filtered rotation angle-radius data sequence to eliminate the influence of high-frequency noise. Specifically, the sampling frequency is set to f. s The rotation angle-radius data sequence after Hampel filtering is transformed from the time domain to the frequency domain using the Fast Fourier Transform (FFT). The FFT is shown in the following equation:

[0052]

[0053] Then set the predetermined cutoff frequency to ω s A low-pass filter is used to filter the rotation angle-radius data sequence in the frequency domain, thereby smoothing the edges of the damaged region to eliminate the influence of high-frequency noise. Exemplarily, the processing method further includes using an inverse fast Fourier transform to transform the rotation angle-radius data sequence in the frequency domain to the time domain, thereby constructing a filtered rotation angle-radius data sequence in the time domain. Exemplarily, this is achieved by controlling the Fourier transform sampling frequency f in the filtering parameters. s and the low-pass filter cutoff frequency ω s It can adjust the specific high-frequency noise to be eliminated and its effects.

[0054] Based on the filtered rotation angle-radius data sequence, a filtered non-destructive testing image can be obtained. Subsequently, based on the filtered non-destructive testing image, all pixels within a defined RGB range can be calculated; these pixels represent the damaged area pixels. Then, using a general pixel-based method, the sum of the sizes of all damaged area pixels is calculated, thus obtaining the accurate area of ​​the damaged area. Multiple non-destructive testing images are processed to obtain the accurate damaged area corresponding to each image. Based on the obtained accurate damaged area and the corresponding residual compression strength data, [A] is analyzed. d ,ε comp It can fit the damage area-residual strength curve.

[0055] For example, the processing method according to a preferred embodiment of the present invention is used to process the non-destructive testing images of X850 composite laminates to determine impact damage. First, multiple C-scan non-destructive testing images of an X850 composite laminate after an impact compression test are obtained. The composite laminate test specimen has dimensions of 150mm × 100mm, a ply length of [45 / 0 / -45 / 90]3s, and a nominal ply thickness of 4.44mm. Five impact energies of 20J, 30J, 35J, 40J, and 50J are used, and each energy level is used to perform six impact compression tests on the composite laminate test specimen. A total of 30 sets of C-scan non-destructive testing images and corresponding test data on residual compressive strength are obtained.

[0056] Taking a C-scan non-destructive testing image of impact damage under 40J impact energy as an example, it consists of 1640 rows and 1470 columns of pixels. The impact point is located at row 894 and column 773, that is, the impact point is located at the geometric center of the damage area. In the non-destructive testing image, the red area is the non-damaged area, and the central non-red area is the impact damage area. The actual dimensions of a single pixel are calibrated to be 0.0727mm in height and 0.0684mm in width.

[0057] Subsequently, the nondestructive testing image is processed by programming using the processing method in the preferred embodiment of the present invention. The RGB range of the damage region is set to (R: 0-200, G: 40-255, B: 20-255), and the damage region containing the splitting area that satisfies this range is obtained. At this time, the nondestructive testing image of impact damage under 40J impact energy is as follows: Figure 2 As shown.

[0058] Next, with the rotation angle step size set to 0.1°, the actual distance between the edge point of the damage area and the impact point corresponding to each rotation angle was calculated sequentially, constructing a rotation angle-radius data sequence. The window size and standard deviation range in the Hampel filtering method were set to 90° and 0.5σ, respectively. The Hampel filtering method was applied to process the rotation angle-radius data sequence. The non-destructive testing image and damage area corresponding to the processed rotation angle-radius data sequence are shown below. Figure 3 As shown.

[0059] The sampling frequency was set to 60Hz. A Fast Fourier Transform (FFT) was applied to convert the rotation angle-radius data sequence filtered by the Hampel filtering method to the frequency domain. After filtering with a low-pass filter at a predetermined cutoff frequency of 0.05Hz, the corresponding non-destructive testing image and damaged area of ​​the filtered rotation angle-radius data sequence are shown below. Figure 4 As shown.

[0060] Figure 4 In the damaged area of ​​the nondestructive testing image shown, the adverse effects of fraying were eliminated and edge roughness was optimized through two filtering methods, resulting in a near-circular shape for the damaged area. At this point, the central damaged area comprises 146,601 pixels, and the calculated damaged area based on the size of a single pixel is 729 mm². 2 .

[0061] Thus, the damage area-compressive residual strength data pair corresponding to the non-destructive testing image of impact damage is obtained [A] d ,ε comp ] = [729mm 2 [4246με]. Following the steps described above, other non-destructive testing images from this experiment were processed, resulting in 30 pairs of damage area-compressive residual strength data. These were then fitted using a negative exponential function to obtain the following results: Figure 5 The figure shows the fitted curve of the damaged area versus the remaining strength.

[0062] This invention extracts edge pixels from nondestructive testing images and calculates the actual distance from each rotation angle's corresponding edge pixel to the impact point, forming a rotation angle-radius data sequence for subsequent filtering. Since the pixels corresponding to the split edges are far from the impact center, they appear as outliers in the data sequence. By employing Hampel filtering, the split areas in the image are identified and eliminated by correcting these outliers. Edge roughness manifests as high-frequency noise in the frequency domain. A low-pass filtering method based on Fast Fourier Transform (FFT) smooths the edges of the damaged area by eliminating this high-frequency noise. Furthermore, by addressing the additional splitting error caused by large impact energy and the error from high-frequency edge noise, a more theoretically consistent circular impact damage area and a more reasonable impact damage area result are obtained.

[0063] The beneficial technical effects of the above-described specific embodiments of the present invention are as follows:

[0064] 1. Edge pixels of the non-destructive testing image were extracted, and the actual distance from the edge pixel corresponding to each rotation angle to the impact point was calculated to form a rotation angle-radius data sequence, which can be used for subsequent filtering processing of the data sequence.

[0065] 2. By using the Hampel filtering method to process the data sequence, the split-wire region in the image was identified and eliminated in order to correct outliers, thereby obtaining accurate damage areas and damage area sizes.

[0066] 3. By adopting a low-pass filtering method based on fast Fourier transform, the edges of the damaged area are smoothed in a way that eliminates high-frequency noise, thus solving the edge noise error problem and obtaining a circular impact damage area that is more in line with the theoretical state.

[0067] 4. By obtaining the accurate area of ​​the damaged region and the corresponding compressive residual strength, a fitting curve of the damaged region area versus the residual strength is obtained. This facilitates subsequent prediction of the damaged region area based on the compressive residual strength, or prediction of the damaged region area of ​​the composite laminate after impact damage based on the obtained compressive residual strength.

[0068] While specific embodiments of the present invention have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of the present invention is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but all such changes and modifications fall within the scope of protection of the present invention.

Claims

1. A method for processing non-destructive testing images to determine impact damage in composite laminates, characterized in that, The processing method includes the following steps: Set the RGB range of the pixel color in the damaged area, and obtain the positions of all pixels that meet the range conditions; Using the impact point in the non-destructive testing image as the rotation center, the actual distance between the edge pixel of the damaged area corresponding to each rotation angle and the impact point is calculated according to the predetermined rotation angle step size, and used as the radius corresponding to that rotation angle, thereby constructing a rotation angle-radius data sequence; The Hampel filtering method is used to identify outliers in the rotation angle-radius data sequence that exceed the allowable standard deviation range, so as to eliminate the splitting areas with abnormal radius in the damaged area; Based on the filtered rotation angle-radius data sequence, the filtered non-destructive testing image is obtained.

2. The processing method as described in claim 1, characterized in that, The impact point is the geometric center of the damage area obtained by the positions of all pixels within the damage area in the non-destructive testing image.

3. The processing method as described in claim 1, characterized in that, The steps for calculating the actual distance between the edge pixel corresponding to each rotation angle and the impact point include: The non-destructive testing image is sized to obtain the actual size corresponding to each pixel. Calculate the pixel distance between the edge pixel and the impact point, and obtain the actual distance by multiplying the pixel distance by the actual size.

4. The processing method as described in claim 1, characterized in that, The steps for identifying outliers using the Hampel filtering method include: Define the window range and construct the window data sequence corresponding to each data point in the rotation angle-radius data sequence; The deviation factor for each data point is calculated based on the window data sequence and the rotation angle-radius data sequence; and If the deviation factor exceeds the allowable standard deviation range, the data point corresponding to the deviation factor is determined to be an outlier.

5. The processing method as described in claim 4, characterized in that, The steps for identifying outliers using the Hampel filtering method further include: Based on the inverse Gaussian error function, the deviation factor corresponding to each data point in the rotation angle-radius data sequence is obtained by using the absolute value of the difference between each data point in the rotation angle-radius data sequence and the median of the corresponding window data sequence, as well as the median of the difference data sequence of the window data sequence.

6. The processing method as described in claim 5, characterized in that, The steps for identifying outliers using the Hampel filtering method further include: Calculate the absolute value of the difference between each data point in each window data sequence and its median to obtain the difference data sequence.

7. The processing method as described in claim 4, characterized in that, The steps for identifying outliers using the Hampel filtering method further include: When the data point corresponding to the deviation factor is determined to be an outlier, the data of the outlier is replaced with the median of the window data sequence corresponding to the data of the outlier, so as to eliminate the splitting area with abnormal radius in the damaged area.

8. The processing method as described in claim 1, characterized in that, The processing method further includes: The filtered rotation angle-radius data sequence is processed using a fast Fourier transform low-pass filtering method to eliminate the influence of high-frequency noise.

9. The processing method as described in claim 8, characterized in that, The steps for filtering using the aforementioned Fast Fourier Transform low-pass filtering method include: By setting a preset sampling frequency, the rotation angle-radius data sequence processed by the Hampel filtering method is transformed from the time domain space to the frequency domain space; A low-pass filter with a predetermined cutoff frequency is used to perform low-pass filtering on the rotation angle-radius data sequence in the frequency domain, thereby smoothing the edges of the damaged region to eliminate the influence of high-frequency noise.

10. The processing method as described in claim 9, characterized in that, The steps of filtering using the aforementioned fast Fourier transform low-pass filtering method further include: The inverse fast Fourier transform is used to transform the rotation angle-radius data sequence after eliminating high-frequency noise to the time domain, thereby constructing the filtered rotation angle-radius data sequence.

11. The processing method as described in claim 1, characterized in that, The processing method further includes: Based on the filtered non-destructive testing image, all pixels within the set RGB range are calculated, and these pixels are the damaged area pixels. The sum of the dimensions of all damaged area pixels is calculated to obtain the area of ​​the damaged area.

12. The processing method as described in claim 11, characterized in that, The processing method further includes: Multiple nondestructive testing images are processed to obtain the damage area corresponding to each nondestructive testing image. Based on the obtained damage area and its corresponding compressive residual strength data, a damage area-residual strength curve is fitted.

Images