A method for detecting defect types of metal components

Abstract

The present application provides a method for detecting defect types of metal components, including: preparing defective samples of metal components with different defect types and defect-free samples of metal components to construct a sample set; emitting pulsed lasers at multiple specific positions of each sample, and simultaneously collecting plasma images and LIBS spectra of the sample; extracting plasma image features and LIBS spectrum features respectively for the plasma images and LIBS spectra collected at the same specific position of the sample to obtain fusion features at each specific position of the sample, and the fusion feature trend of the sample; constructing a defect recognition model based on the fusion feature trend and defect type of each sample using a linear discriminant analysis method; collecting plasma images and LIBS spectra at multiple target positions of the metal component to be detected to obtain a fusion feature trend, and detecting the defect type using a defect recognition model. The method can improve the detection performance of defect types by combining plasma image features.

Description

Technical Field

[0001] This application belongs to the technical field of metal component defect detection, and in particular, relates to a method for detecting the type of metal component defects. Background Technology

[0002] Metal additive manufacturing (AM) technology has been widely used in aerospace, automobile manufacturing and other industries in recent years. However, due to its complex processing process, AM components often have defects such as pores, cracks, and protrusions, which may affect the performance and reliability of the components. Therefore, the detection technology for AM component defects has become a hot topic in current research.

[0003] As a high-precision and high-sensitivity optical analysis method, laser-induced breakdown spectroscopy (LIBS) technology is widely used for rapid detection of element types and concentrations. With the in-depth study of the relationship between LIBS spectral characteristics and element types, the application of this technology has expanded from traditional component analysis to more complex fields, such as circuit aging prediction, battery evaluation, and material performance monitoring. Especially in the field of metal AM, LIBS technology has shown great potential in defect detection due to its significant advantages of no sample preparation, rapid in-situ detection, and high spatial resolution. Therefore, many domestic and foreign researchers have applied LIBS to defect identification and quality control of metal AM components.

[0004] In the study of spatial effects of LIBS, domestic and foreign scholars mainly focus on enhancing the signal intensity of plasma and improving experimental repeatability through different spatial constraint structures and sizes. For cylindrical cavities, N 2 The dual enhancement mechanism of atmosphere and cylindrical cavity confinement is used to improve the detection performance of LIBS for Sr elements in soil. Under the optimal conditions (cavity diameter 2mm, height 6mm), the enhancement factor is 3.2. For the hemispherical cavity, the effect of laser energy on the enhancement characteristics of the hemispherical cavity confined plasma is studied. When the laser energy exceeds 80mJ, the plasma under the hemispherical cavity confinement is significantly enhanced within 8-14μs, and the enhancement factor increases with the increase of laser energy. For the parallel wall cavity, by synchronously acquiring the plume and shock wave images, it is found that the reflected shock wave under the parallel cavity confinement compresses the plasma, significantly enhances the emission intensity and improves the signal reproducibility, and this effect only appears within a specific time window.

[0005] Most of the above studies are based on the ablation point of LIBS acting on the central position of symmetric structures (such as cylindrical cavities, hemispherical cavities, and parallel wall cavities), forming symmetric spatial effects, so that the reflected shock waves can uniformly compress the plasma. However, in actual metal AM components, defects often exhibit asymmetric structures, resulting in asymmetric characteristics of the spatial effects caused by the defects. At present, there is no report on the interaction mechanism of asymmetric spatial effects on LIBS plasma. Summary of the Invention

[0006] In view of this, the purpose of this application is to provide a method for detecting the defect types of metal components. By combining plasma image features and LIBS spectral features, the detection performance of defect types can be improved. In particular, based on the asymmetric spatial effects exhibited by crack defects, crack defects and pit defects can be fully distinguished.

[0007] This application provides a method for detecting the defect types of metal components, and the method includes:

[0008] Preparing metal component defect samples and metal component defect-free samples with different defect types to construct a sample set; wherein, the defect types include: cracks, pits, and protrusions;

[0009] For each sample, emitting pulsed lasers to multiple specific positions of the sample, and simultaneously collecting the plasma image and LIBS spectrum of the sample;

[0010] For the plasma image and LIBS spectrum collected at the same specific position of the sample, respectively extracting plasma image features and LIBS spectral features to obtain the fusion features at each specific position of the sample;

[0011] Based on the fusion features at each specific position of the sample, obtaining the fusion feature trend of the sample;

[0012] Based on the fusion feature trends and defect types of each sample, using linear discriminant analysis to construct a defect recognition model;

[0013] Collecting the plasma image and LIBS spectrum at multiple target positions of the metal component to be detected to obtain the fusion feature trend of the metal component to be detected, and using the defect recognition model to detect the defect type of the metal component to be detected.

[0014] Further, when the sample is a metal component defect sample, the emitting pulsed lasers to multiple specific positions of the sample and simultaneously collecting the plasma image and LIBS spectrum of the sample includes:

[0015] Determining the defect position of the sample and identifying the defect edge;

[0016] Starting from any edge of the defect and ending at the approximate center position of the defect, multiple specific positions are selected at specific length intervals; wherein, the specific positions do not include the defect edge.

[0017] For each specific position, pulsed laser ablation is performed, and at the same time, the plasma image and LIBS spectrum of the sample at this specific position are collected.

[0018] Further, when the sample is a defect-free sample of a metal component, the pulsed laser ablation is performed at multiple specific positions of the sample, and at the same time, the plasma image and LIBS spectrum of the sample are collected, including:

[0019] Multiple specific positions are randomly selected on the surface of the sample.

[0020] For each specific position, pulsed laser ablation is performed, and at the same time, the plasma image and LIBS spectrum of the sample at this specific position are collected.

[0021] Further, before separately extracting the plasma image features and LIBS spectrum features to obtain the fusion features at each specific position of the sample, the method further includes:

[0022] For each plasma image, the edge detection algorithm is used to divide the plasma image into different regions to obtain plasma image features.

[0023] For each LIBS spectrum, the principal component analysis method is used to extract the principal components that can reflect the spectrum features to obtain LIBS spectrum features.

[0024] Further, the using the edge detection algorithm to divide the plasma image into different regions to obtain plasma image features includes:

[0025] The edge detection method is used to divide the plasma image from the inside to the outside into internal, middle and external regions.

[0026] The boundary line between the internal and middle regions is defined as the middle-inner layer, and the boundary line between the middle and external regions is defined as the middle-outer layer.

[0027] Using image processing technology, the contour lengths and in-layer areas of the middle-inner layer and the middle-outer layer are extracted as key dimension features.

[0028] Using the key dimension features, the near-circularity ratios of the middle-inner layer and the middle-outer layer are respectively calculated, as well as the proportions of the middle and external regions in the plasma image, so as to obtain plasma image features.

[0029] Further, extracting the principal components that can reflect the spectral characteristics by using the principal component analysis method to obtain the LIBS spectral characteristics includes:

[0030] Performing preprocessing of background subtraction and normalization on the LIBS spectrum;

[0031] Based on the preprocessed LIBS spectrum, extracting multiple principal components that can explain the LIBS spectral characteristics and calculating the explained variance of each principal component;

[0032] According to the cumulative proportion of the explained variance, selecting a specific number of top-ranked principal components to construct the LIBS spectral characteristics.

[0033] Further, collecting plasma images and LIBS spectra at multiple target positions of the metal component to be detected to obtain the fusion feature trend of the metal component to be detected includes:

[0034] Extracting the plasma image features and LIBS spectral features at multiple target positions of the metal component to be detected to obtain the fusion features at multiple target positions of the metal component to be detected;

[0035] Based on the fusion features at multiple target positions of the metal component to be detected, obtaining the fusion feature trend of the metal component to be detected.

[0036] The metal component defect type detection method provided by this application can improve the detection performance of defect types by combining plasma image features and LIBS spectral features. In particular, based on the asymmetric spatial effect exhibited by crack defects, crack defects and pit defects can be fully distinguished. Description of the Drawings

[0037] Figure 1 Shows a schematic diagram of a parallel-wall cavity sample provided by an embodiment of this application;

[0038] Figure 2 Shows a collection device for plasma images and LIBS spectra provided by an embodiment of this application;

[0039] Figure 3 Shows a graph of the evolution trend of spectral intensity over time under unconstrained conditions provided by an embodiment of this application;

[0040] Figure 4 Shows a schematic diagram of the plasma evolution process of the asymmetric spatial effect provided by an embodiment of this application;

[0041] Figure 5 Shows a graph of the change trends of spectral intensity, delay time, and enhancement factor under the reflection shock waves of the left and right sidewalls of the parallel cavity provided by an embodiment of this application;

[0042] Figure 6 Shows the evolution images of the plasma morphology under asymmetric space constraints (DAPLW = 1mm - 7mm) and without space constraints provided by the embodiments of the present application;

[0043] Figure 7 Shows the flow chart of the method for detecting the defect types of metal components provided by the embodiments of the present application;

[0044] Figure 8 Shows the example diagrams of the defective samples and non - defective samples of metal components provided by the embodiments of the present application;

[0045] Figure 9 Shows the regional division model diagram of the plasma image provided by the embodiments of the present application. Detailed implementation manners

[0046] To make the purpose, technical solutions, and advantages of the technical solution more clear and understandable, the technical solution will be further described in detail below in combination with specific implementation manners. It should be understood that these descriptions are only exemplary and do not limit the scope of the technical solution.

[0047] Embodiment 1

[0048] Embodiment 1 of the present application explores the influence of the asymmetric space effect on the LIBS plasma morphology and spectral characteristics. First, using ferrochrome alloy materials, a planar sample and a parallel - wall cavity sample as shown in Figure 1 are prepared by metal additive manufacturing technology to simulate crack defects in metal components. Among them, Figure 1 (a), (b), and (c) are the three - dimensional view, top view, and front view of the parallel - wall cavity sample respectively. As shown in Figure 1 , the outer surface size of the parallel - wall cavity is 24mm×15mm×5mm, and the width of the inner surface of the parallel - wall cavity is 14mm and the height is 3mm. Here, the defect types affecting the quality of metal components mainly include: cracks, pits, and protrusion defects. Among them, protrusion defects are relatively easy to identify, while cracks and pit defects are difficult to distinguish because they both present a concave surface. Therefore, clarifying the spectral characteristics and plasma behavior characteristics of one of the defects means distinguishing between the two defects. Therefore, in Embodiment 1 of the present application, the crack defect with more obvious spectral characteristics and plasma behavior characteristics is used as the research object. Further, the parallel - wall cavity similar to the crack defect is used as the experimental model.

[0049] Secondly, build as shown in Figure 2The acquisition device for the plasma image and LIBS spectrum shown. A Q-switched Nd:YAG laser (model Vlite-200, wavelength 1064 nm, energy 60 mJ, pulse width 8 ns, repetition rate 1 Hz, beam mode TEM00) is used as the ablation source. The pulsed laser is reflected by mirrors and lenses and focused 2 mm below the sample surface, inducing a plasma. The optical signal emitted by the plasma is collected by an optical fiber probe, and the signal is transmitted to a spectrometer (SR-500i-a, Andor). Subsequently, an ICCD detector (1024×1024 pixels, Andor) converts the optical signal into digital information and stores it in a computer. The spectral wavelength acquisition range is 250 nm to 650 nm, and the resolution is 0.07 nm. The gate width is set to 1 μs, and the delay step is 1 μs to measure the emission time characteristics of the plasma during the acquisition delay from 1 μs to 22 μs. At the same time, the plasma image is recorded by a high-speed camera (model Dicam-pro, zoom range 24 - 85 mm, aperture range f / 22 to f / 2.8, resolution 1280x1024, frame rate 8 fps, pixel size 6.7 μm), and the image data is saved to the computer for subsequent analysis. The acquisition processes of the LIBS spectrum information and the plasma image information are carried out synchronously, and the timing of all behaviors such as laser emission and plasma information acquisition is regulated by a multi-functional digital delay pulse generator (model BNC-575).

[0050] Then, for the planar sample, the pulsed laser is directly irradiated on the sample surface. For the parallel-wall cavity sample, the distance between the ablation point and the left parallel wall is defined as DAPLW, and DAPLW is made to be 1 mm, 2 mm, 3 mm, 4 mm, 5 mm, 6 mm, 7 mm (located in the middle part of the groove) respectively, where 7 mm is the symmetric position and the rest are asymmetric positions.

[0051] Finally, spectral characteristic analysis and plasma behavior analysis are carried out on the acquired LIBS spectrum and plasma image.

[0052] (1) Select the characteristic spectral lines of the LIBS spectrum:

[0053] The spectral lines selected in the first embodiment of this application should meet the following principles: They must still be effectively acquired in the later stage of the plasma to ensure the stability and reliability of data during actual measurement; they should be less interfered by other elements, which can reduce the influence of cross-interference on the spectral line intensity, thereby improving the analysis accuracy. Here, please refer to the trend graph of the spectral intensity versus time under unconstrained conditions shown in Figure 3 where, Figure 3 (a) is the overall evolution trend within the delay time range of 1 - 22 μs, Figure 3(b) is the spectrogram when the delay time is 22 μs. As shown in the figure, according to the above two principles, three spectral lines, Cr I 425.43 nm, Mn II 463.92 nm, and Fe I 519.87 nm, are finally determined as the characteristic spectral lines of the first embodiment of this application.

[0054] (2) Theoretical analysis: When the laser-induced plasma expands in air, the plasma from the ablated target will generate a shock wave, which will affect the entire LIBS process. Specifically, based on the spatial confinement effect, the increase in plasma temperature and density is caused by the reflection of the shock wave from the cavity wall. When the shock wave reaches the inner surface of the confinement cavity and reflects back, it continues to propagate towards the plasma, interacts with the plasma, enhances the particle collisions, and thus increases the plasma temperature and density. At the same time, the shock wave heating effect also enhances the emission intensity of the plasma, thereby increasing the intensity of the spectral lines. For the plasma evolution process with asymmetric spatial effects, the ablation point is located at an asymmetric position, and a schematic diagram of the plasma evolution process with asymmetric spatial effects as shown in Figure 4 is drawn. When the plasma is induced at an asymmetric position (to the left of the center line), the generation of the plasma is accompanied by the outward diffusion of the front shock wave (see Figure 4 (a)); this diffused shock wave is reflected when it encounters the closer left wall, forming a left wall reflected shock wave (see Figure 4 (b)); this left wall reflected shock wave acts on the plasma first, causing changes in its spectral characteristics and morphology (see Figure 4 (c)); subsequently, the plasma is further affected by the right wall reflected shock wave, thus triggering more complex changes (see Figure 4 (d)).

[0055] (3) Analyze spectral characteristics:

[0056] In the first embodiment of this application, the following parameters are quantitatively analyzed: spectral intensity, enhancement factor, and corresponding delay time to explore the spectral characteristics of defective samples. Among them, the spectral intensity is the most intuitive spectral characteristic parameter, which can provide direct information about the plasma state; the enhancement factor quantifies the degree of improvement of the spectral intensity under the asymmetric spatial effect relative to the unconstrained spectral emission signal, reflecting the influence of different DAPLW on the plasma characteristics; the corresponding delay time is used to describe the response time of the excitation signal corresponding to the enhancement factor.

[0057] Explore the influence of asymmetric spatial effects on spectral characteristics: Please refer to as shown in Figure 5The trend charts of spectral intensity, delay time, and enhancement factor under the reflection of shock waves on the left and right sidewalls of the parallel cavity are shown. Among them, when DAPLW is 1 mm - 6 mm, asymmetric spatial confinement is generated, and when DAPLW is 7 mm, symmetric spatial confinement is generated.

[0058] From Figure 5 the overall change trend of the spectral intensity shown in (a) and (d), it can be seen that as DAPLW increases, the spectral intensity of the characteristic spectral lines under the reflection of the shock wave on the left sidewall shows a downward trend; while the spectral intensity under the reflection of the shock wave on the right sidewall shows an upward trend. From Figure 5 the overall change trend of the delay time shown in (b) and (e), it can be seen that the initial action delay time of the shock wave reflected from the left sidewall is 2 µs (DAPLW = 1 mm), and the propagation time increases with the increase of DAPLW; while the initial action delay time of the shock wave reflected from the right sidewall is 16 µs (DAPLW = 3 mm). This is because when DAPLW is 1 and 2 mm, the shock wave energy is weak when the shock wave reflected from the right sidewall acts on the plasma, and the enhancement effect is not obvious, and the propagation time decreases with the increase of DAPLW; finally, the synchronous action time of the shock waves reflected on both sides is 10 µs (DAPLW = 7 mm). From Figure 5 the overall change trend of the enhancement factor shown in (c) and (f), it can be seen that all enhancement factors are greater than 1, indicating that the shock waves reflected on both the left and right sides have an enhancing effect on the characteristic spectral lines. As DAPLW increases from 2 mm to 7 mm, the enhancement effect of the shock wave reflected from the left sidewall on the spectral line gradually decreases and reaches the lowest point at DAPLW = 6 mm; it rebounds at DAPLW = 7 mm due to the influence of the shock wave reflected from the right sidewall. In contrast, the enhancement effect of the shock wave reflected from the right sidewall on the spectral line shows a gradually increasing trend.

[0059] The above phenomena indicate that when DAPLW changes from 1 to 6 mm, there are significant differences in the effects of the shock waves reflected on the left and right sidewalls on the spectrum, that is, an asymmetric spatial effect appears. It should be noted that when DAPLW = 7 mm, the shock waves reflected on both the left and right sidewalls act on the plasma synchronously, and at this time, the symmetric spatial confinement effect appears, and the enhancement factors of the characteristic spectral lines reach 4.5, 4.71, and 3.73 respectively. This shows that under the symmetric spatial effect, the enhancement effect of the spectral line intensity is better than that of the asymmetric spatial effect. Specifically, the asymmetric spatial effect will cause the plasma to be unable to be compressed evenly, resulting in partial loss of energy, and further affecting the spectral line intensity. Under the symmetric spatial effect, the plasma can be compressed more evenly, and the spectral intensity is enhanced more effectively.

[0060] (4) Analyze the plasma behavior characteristics:

[0061] Please refer to as Figure 6Evolution images of the plasma morphology under the shown asymmetric spatial constraint (DAPLW = 1 mm - 7 mm) and without spatial constraint. The first image under all conditions was taken with a time delay of 1 μs, and the time delay of each subsequent image increased by 1 μs. As can be seen from Figure 6 it that the evolution of the plasma morphology can be roughly divided into four stages:

[0062] Stage 1 (1 μs): The plasma is relatively small, but has a high temperature and electron density, resulting in a large spectral intensity.

[0063] Stage 2 (2 μs - 10 μs): The plasma morphology is relatively stable. In particular, when DAPLW = 1 mm, the plasma shows a tendency to tilt to the right, while there is no significant tilting change under other conditions. This is because when DAPLW = 1 mm, the ablation point is closer to the left wall, and the reflected shock wave energy compresses the plasma, causing an external force on the plasma towards the right wall, thus resulting in a tendency to shift to the right. As DAPLW increases, the energy of the shock wave reflected by the left wall gradually weakens, and the rightward deviation tendency of the plasma also gradually decreases. When DAPLW = 2 mm - 7 mm, there is no obvious tilting change in the plasma morphology.

[0064] Stage 3 (10 μs - 13 μs): The plasma morphology changes significantly under all conditions, which is due to the change in the particle density core of the plasma. At this time, the evolution trend of the plasma with DAPLW = 1 mm - 7 mm is similar to the natural evolution trend under the condition of no spatial constraint.

[0065] Stage 4 (13 μs - 22 μs): As time goes by, the temperature and electron density of the plasma gradually decrease, and the morphology is more vulnerable to external factors. For example, under the condition of DAPLW = 3 mm, when the delay time is 15 μs, the plasma morphology is affected by the compression of the shock wave reflected by the right wall and the shock wave energy brought in, resulting in morphological distortion, which affects the temperature and density of the plasma and causes changes in the spectral intensity. Similar phenomena also occur under the conditions of DAPLW = 4 mm - 6 mm. However, under the conditions of DAPLW = 1 mm and 2 mm, although similar changes occurred in the plasma morphology, they were not reflected in the spectral characteristics. This may be because the shock wave energy reflected by the right reflection wall is relatively low. Although it affects the plasma morphology, it is not sufficient to cause significant changes in temperature and density.

[0066] By comparing the changes in the plasma morphology under different conditions, the following conclusions are drawn: The plasma morphology under the conditions of no spatial constraint and symmetric spatial constraint (DAPLW = 7 mm) is relatively stable. The reason is that under the condition of no spatial constraint, the plasma is not disturbed by external forces, and under the action of symmetric spatial constraint, the reflected shock wave uniformly compresses the plasma, resulting in relatively small changes in the plasma morphology. That is to say, the evolution process of the plasma is significantly affected by the asymmetric spatial constraint. Especially at a small DAPLW value, the action of the reflected shock wave significantly changes the morphology and spectral characteristics of the plasma.

[0067] The evolution of the plasma morphology can reflect the microscopic structure or morphological changes on the material surface. In the defect area, the formation and evolution of the plasma may exhibit characteristics different from those in the normal area, such as irregular morphology and uneven expansion. By monitoring the evolution trend of the plasma morphology, especially the local irregular changes caused by defects, it can assist the LIBS technology in effectively detecting defects. Further, the evolution process of the plasma morphology can be characterized by the near-circularity of the plasma.

[0068] In summary, compared with the performance under the symmetric spatial effect and the unconstrained condition, the plasma under the asymmetric spatial constraint exhibits different behavioral characteristics. This discovery is of great significance for applying the LIBS technology to fields with asymmetric structural characteristics (such as metal defect detection).

[0069] Example Two

[0070] Embodiment Two of the present application discloses a method for detecting the types of defects in metal components. Please refer to Figure 7 the flowchart of the method for detecting the types of defects in metal components provided in the embodiment of the present application as shown. The method includes:

[0071] S101. Prepare metal component defect samples with different defect types and metal component defect-free samples to construct a sample set.

[0072] Among them, the defect types include cracks, pits, and protrusions.

[0073] In this step, the number of prepared crack, pit, and protrusion defect samples and defect-free samples is the same. Here, for the convenience of measurement and modeling, the sizes of the three types of defects, that is, the width of the crack and the diameters of the pit and the protrusion, can be controlled to be of a unified size. As an example, this size can be set at about 2 mm. Here, please refer to Figure 8 the example diagrams of the metal component defect samples and metal component defect-free samples as shown.

[0074] S102. For each sample, emit pulsed laser at multiple specific positions of the sample, and simultaneously collect the plasma image and LIBS spectrum of the sample.

[0075] In this step, pulsed laser is emitted at specific positions of the sample for ablation to excite plasma, so as to simultaneously collect the plasma image and LIBS spectrum of the sample using the acquisition device for plasma image and LIBS spectrum as shown in Figure 2 Figure.

[0076] In addition, according to the variation trend of the spectral characteristics and plasma behavior characteristics with the delay time described in Embodiment 1, in Embodiment 2 of the present application, the delay time is set to 2 μs to avoid the delay time having a greater impact on the spectral characteristics and plasma behavior characteristics at different specific positions, so as to regard the asymmetric spatial effect as the main influencing factor causing the change of spectral intensity and plasma morphology, and at the same time, by setting different acquisition positions, the plasma images and LIBS spectra under different asymmetric spatial effects are obtained.

[0077] Specifically, when the sample is a metal component defect sample, the emitting pulsed laser at multiple specific positions of the sample and simultaneously collecting the plasma image and LIBS spectrum of the sample includes:

[0078] Step 1021. Determine the defect position of the sample and identify the defect edge.

[0079] Step 1022. Starting from any edge of the defect and ending at the approximate center position of the defect, select multiple specific positions at a specific length interval.

[0080] Among them, the specific positions do not include the defect edge.

[0081] In this step, the length interval can be set according to the size of the defect to determine multiple specific positions, so as to prepare for obtaining the plasma images and LIBS spectra under multiple different asymmetric spatial effects. As an example, the length interval in Embodiment 2 of the present application can be set to 250 μm. Further, 4 specific positions can be determined, and the distances between the specific positions and the defect edge are 250 μm, 500 μm, 750 μm, and 1 mm respectively. It should be noted that the specific positions should be selected inside the defect rather than at the defect edge. In addition, for crack defect samples, specific positions need to be set along the width direction of the crack.

[0082] Step 1023. For each specific position, emit pulsed laser for ablation and simultaneously collect the plasma image and LIBS spectrum of the sample at this specific position.

[0083] When the sample is a defect-free metal component sample, emitting pulsed laser to ablate at multiple specific positions of the sample, and simultaneously collecting the plasma image and LIBS spectrum of the sample, including:

[0084] Step 1024: Randomly select multiple specific positions on the surface of the sample.

[0085] In this step, in order to maintain data consistency, the length intervals of the defect-free sample and the defective sample are made the same, so that the number of specific positions is the same.

[0086] Step 1025: For each specific position, emit pulsed laser to ablate, and simultaneously collect the plasma image and LIBS spectrum of the sample at this specific position.

[0087] S103: For the plasma image and LIBS spectrum collected at the same specific position of the sample, extract the plasma image features and LIBS spectrum features respectively to obtain the fusion features at each specific position of the sample.

[0088] In this step, the plasma image features are used to characterize the plasma behavior characteristics, and the LIBS spectrum features are used to characterize the spectral characteristics. Based on the analysis results of the above-mentioned Embodiment 1, at the same specific position, the plasma behavior characteristics and spectral characteristics of the sample are consistent. Then, fusing the plasma image features and LIBS spectrum features can enhance the feature representation at this specific position, that is, obtain the fusion features at this specific position.

[0089] As an example, obtain the plasma image and LIBS spectrum collected at a specific position with a distance of 250 µm from the edge, so as to extract the plasma image features and LIBS spectrum features corresponding to this specific position.

[0090] Before respectively extracting the plasma image features and LIBS spectrum features to obtain the fusion features at each specific position of the sample, the method further includes:

[0091] Step 1031: For each plasma image, use an edge detection algorithm to divide the plasma image into different regions to obtain plasma image features.

[0092] In specific implementation, the plasma image features can be obtained in the following way:

[0093] Step 201: Use an edge detection method to divide the plasma image from the inside to the outside into an internal, middle, and external region.

[0094] Step 202: Define the boundary line between the internal and middle regions as the middle-inner layer, and the boundary line between the middle and external regions as the middle-outer layer.

[0095] In this step, reference can be made to Figure 9 the regional division model diagram of the plasma image as shown.

[0096] Step 203: Using image processing technology, extract the contour lengths and in-layer areas of the middle inner layer and the middle outer layer as key dimension features.

[0097] Step 204: Using the key dimension features, calculate the near-circularity rates of the middle inner layer and the middle outer layer respectively, as well as the proportions of the middle and outer regions in the plasma image, so as to obtain plasma image features.

[0098] Step 1032: For each LIBS spectrum, use the principal component analysis method to extract the principal components that can reflect the spectrum features to obtain LIBS spectrum features.

[0099] In specific implementation, the LIBS spectrum features can be obtained through the following method:

[0100] Step 301: Perform preprocessing of background subtraction and normalization on the LIBS spectrum.

[0101] Step 302: Based on the preprocessed LIBS spectrum, extract multiple principal components that can explain the LIBS spectrum features and calculate the explained variance of each principal component.

[0102] Step 303: According to the cumulative proportion of the explained variance, select a specific number of principal components with the top rankings to construct LIBS spectrum features.

[0103] As an example, the top 3 principal components, that is, the characteristic spectral lines Cr I 425.43nm, Mn II 463.92nm, and Fe I 519.87nm selected in the first embodiment above, are used as the principal components of the LIBS spectrum features to construct LIBS spectrum features.

[0104] S104: Based on the fusion features at each specific position of the sample, obtain the fusion feature trend of the sample.

[0105] In this step, different specific positions correspond to different feature representations, that is, asymmetric spatial effects or symmetric spatial effects. Based on the different feature representations at each specific position, obtain the change trend of the fusion features of the sample with each specific position. For different defective samples, the change trend of the fusion features with each specific position is also different, that is, the fusion feature trend can characterize the defect type of the sample.

[0106] S105: Based on the fusion feature trends and defect types of each sample, use the linear discriminant analysis method to construct a defect recognition model.

[0107] In this step, based on the fusion feature trends of various samples, linear discriminant analysis is used for classification to obtain the predicted defect types of various samples. The parameters of the linear discriminant analysis are adjusted based on the original defect types of various samples so that the predicted defect types of various samples tend to the original defect types of various samples, thereby obtaining a trained defect recognition model.

[0108] S106. Collect plasma images and LIBS spectra at multiple target positions of the metal component to be detected to obtain the fusion feature trend of the metal component to be detected, and use the defect recognition model to detect the defect type of the metal component to be detected.

[0109] In specific implementation, the fusion feature trend of the metal component to be detected can be obtained through the following method:

[0110] Step 1061. Extract plasma image features and LIBS spectral features at multiple target positions of the metal component to be detected to obtain the fusion features at multiple target positions of the metal component to be detected.

[0111] In this step, since the defect type and defect position of the metal component to be detected are unknown, multiple target positions can be selected on the surface of the metal component to be detected according to the length interval when selecting specific positions, a target position sequence is constructed, plasma images and LIBS spectra are collected at each target position, and plasma image features and LIBS spectral features are extracted to obtain the fusion features at multiple target positions.

[0112] Step 1062. Based on the fusion features at multiple target positions of the metal component to be detected, obtain the fusion feature trend of the metal component to be detected.

[0113] In this step, the number of the above selected specific positions is used as the target number, the target number of target positions are extracted overlapped from the target position sequence, and the fusion feature trend is obtained based on the fusion features of the extracted target positions until the target position sequence is traversed.

[0114] The above content is only the preferred embodiment of the present invention. For those of ordinary skill in the art, many changes can be made in the specific implementation manner and application scope according to the idea of the present technical content. As long as these changes do not depart from the concept of the present invention, they all belong to the protection scope of this patent.

Claims

1. A method for detecting defect types of metal components, characterized in that: The method comprises: Preparing defective samples of metal components with different defect types and non-defective samples of metal components to construct a sample set; wherein the defect types include: cracks, potholes and bulges; For each sample, a pulsed laser is emitted to multiple positions of the sample, and the plasma image and LIBS spectrum of the sample are collected simultaneously; For the plasma image and LIBS spectrum collected at the same position of the sample, the plasma image features and the LIBS spectrum features are extracted respectively to obtain the fusion features at each position of the sample; Based on the fusion features at each position of the sample, a fusion feature trend of the sample is obtained; Based on the fusion feature trends and defect types of each sample, a defect recognition model is constructed using linear discriminant analysis; The plasma images and LIBS spectra at multiple target positions of the metal component to be detected are collected to obtain the fusion feature trend of the metal component to be detected, and the defect type of the metal component to be detected is detected using the defect recognition model.

2. The method according to claim 1, characterized in that When the sample is a metal component defect sample, emitting pulsed lasers at multiple positions of the sample and simultaneously collecting a plasma image and a LIBS spectrum of the sample include: Determine the defect location of the sample and identify the defect edge; Taking any edge of the defect as a starting point and the center of the defect as an end point, setting a length interval according to the size of the defect to select a plurality of first positions; wherein the first position does not include the defect edge; For each first position, a pulsed laser is emitted for ablation, and a plasma image and a LIBS spectrum of the sample at the first position are simultaneously collected.

3. The method according to claim 1, characterized in that When the sample is a defect-free metal component sample, emitting pulsed lasers at multiple positions of the sample for ablation, and simultaneously collecting a plasma image and a LIBS spectrum of the sample, comprises: A plurality of second positions are randomly selected on the surface of the sample; For each second position, a pulsed laser is emitted to perform ablation, and a plasma image and a LIBS spectrum of the sample at the second position are simultaneously collected.

4. The method according to claim 1, characterized in that Before respectively extracting the plasma image features and the LIBS spectrum features to obtain the fusion features at each position of the sample, the method further comprises: For each plasma image, the plasma image is divided into different regions using an edge detection algorithm to obtain plasma image features; For each LIBS spectrum, the principal component analysis method is used to extract the principal component that can reflect the spectral characteristics to obtain the LIBS spectral characteristics.

5. The method according to claim 4, characterized in that The plasma image is divided into different regions using an edge detection algorithm to obtain plasma image features, including: The plasma image is divided into inner, middle and outer regions from the inside to the outside by using an edge detection method; The dividing line between the inner and middle regions is defined as the middle-inner layer, and the dividing line between the middle and outer regions is defined as the middle-outer layer; By using image processing technology, the contour length and the inner area of ​​the middle inner layer and the middle outer layer are extracted as key size features; The key size features are used to calculate the near-circularity of the middle inner layer and the middle outer layer, as well as the proportion of the middle and outer regions in the plasma image, so as to obtain the plasma image features.

6. The method according to claim 4, characterized in that The principal component analysis method is used to extract the principal component that can reflect the spectral characteristics to obtain the LIBS spectral characteristics, including: Preprocessing of LIBS spectra for background subtraction and normalization; Based on the preprocessed LIBS spectra, multiple principal components that can explain the LIBS spectral characteristics are extracted, and the explained variance of each principal component is calculated; According to the cumulative proportion of explained variance, a specific number of principal components with high rankings are selected to construct LIBS spectral features; wherein the specific number is 3.

7. The method according to claim 1, characterized in that The collecting of plasma images and LIBS spectra at multiple target positions of the metal component to be detected to obtain a fusion feature trend of the metal component to be detected includes: Extracting plasma image features and LIBS spectrum features at multiple target positions of the metal component to be detected to obtain fusion features at multiple target positions of the metal component to be detected; Based on the fusion features at multiple target positions of the metal component to be detected, a fusion feature trend of the metal component to be detected is obtained.

Images