Metal material defect identification and analysis method and system based on multiple detection technology

Abstract

The present invention provides a metal material defect recognition and analysis method and system based on multiple detection technology, which relates to the field of metal defect detection technology. The method comprehensively utilizes microscope detection and ultrasonic detection technology to obtain surface morphology data and internal structure data of the target metal material. The two types of data are feature aligned and fused by a multiple grid method to generate a unified feature distribution model, and a mathematical solution model for ultrasonic wave field propagation is established based on this model. The location, shape and size information of potential defects are determined using the solution data, and defect location data containing three-dimensional coordinates and contour features are generated. The key defect information is extracted in combination with the edge detection algorithm, and the defect type is determined by the classification algorithm to generate the detection result. The present invention helps to solve the problems of insufficient detection accuracy, difficult data fusion and low classification efficiency in the prior art, and improves the applicability and reliability of defect detection technology.

Description

Technical Field

[0001] The present invention relates to the technical field of metal defect detection, and in particular to a metal material defect recognition and analysis method and system based on multiple detection technologies. Background Art

[0002] Metal materials will inevitably produce surface or internal defects during production and use, which may lead to product performance degradation or even safety accidents. Therefore, high-precision nondestructive testing technology for metal materials has become a research hotspot.

[0003] For example, the Chinese patent with publication number CN118330030A discloses a method for detecting and imaging internal defects of metal plates. This method utilizes the propagation characteristics of Lamb waves in metal plates and combines the damage probability imaging algorithm based on CEEMDAN (adaptive noise complete set empirical mode decomposition) to effectively detect internal defects of metal plates. The specific method includes: selecting excitation signals according to the dispersion curve of Lamb waves, arranging grids and sensor arrays, generating damage probability imaging maps through signal decomposition and reconstruction and Hilbert transform, and thus locating defects. The improvement of this method is that the anti-noise performance of signal processing is improved, so that the quality of defect imaging is significantly improved.

[0004] However, the above methods have weak comprehensive analysis capabilities for surface defects and internal defects, making it difficult to achieve a comprehensive assessment of complex defects. Summary of the invention

[0005] In view of the deficiencies of the prior art, the present invention provides a metal material defect identification and analysis method and system based on multiple detection technologies.

[0006] In a first aspect, the present invention provides a metal material defect identification and analysis method based on multiple detection technology, comprising:

[0007] Using a preset microscope device to detect the surface morphology of the target metal material to obtain surface morphology data; wherein the surface morphology data includes: image data of the surface of the target metal material, and size, shape and distribution information of surface defects of the target metal material;

[0008] The internal structure of the target metal material is detected based on ultrasonic detection technology to generate ultrasonic detection data; wherein the ultrasonic detection data includes: waveform amplitude, frequency characteristics, depth and shape information of the internal defects of the target metal material;

[0009] Based on the multi-grid method, feature alignment and fusion processing are performed on the surface morphology data and the ultrasonic detection data to generate solution data of ultrasonic wave field propagation; based on the solution data, the location, shape and size information of the potential defect are determined to generate defect location data including three-dimensional coordinates and contour features;

[0010] Based on the defect location data, using an edge detection algorithm, extract key defect information of the target metal material; wherein the key defect information includes the contour and depth information of the defect;

[0011] Based on the key defect information, the defect type of the target metal material is determined, and the detection result information corresponding to the target metal material is generated.

[0012] As an optional implementation, the generating of solution data for ultrasonic wave field propagation includes:

[0013] The surface morphology data and the ultrasonic detection data are respectively subjected to layered processing to extract feature information at different scales;

[0014] Performing alignment processing on the features of the surface topography data and the ultrasonic detection data at each level by using a multi-grid method, wherein the alignment processing includes feature point matching and geometric transformation;

[0015] After completing the alignment of each level, the alignment results of each level are fused through the grid interpolation method to generate a unified feature distribution model;

[0016] According to the characteristic distribution model, a mathematical solution model for ultrasonic wave field propagation is established to generate solution data for ultrasonic wave field propagation.

[0017] As an optional implementation, the multi-grid method includes:

[0018] Based on the geometric constraints of the surface morphology of the target metal material, extracting significant feature points and generating a surface morphology feature point set;

[0019] Generate an internal structure feature point set based on the depth information and waveform amplitude characteristics of the internal structure in the ultrasonic detection data;

[0020] The surface morphology feature point set and the internal structure feature point set are matched at multiple scales through the morphology constraint registration method.

[0021] As an optional implementation, the multi-scale morphology matching includes:

[0022] The surface morphology feature point set is fitted into a continuous surface using a surface fitting algorithm to generate a constrained surface model;

[0023] According to the constrained surface model, a deep feature optimization algorithm is used to transform and adjust the internal structure feature point set to achieve global feature alignment;

[0024] In response to the completion of global feature alignment, the surface morphology features and the ultrasonic detection features are dynamically adjusted based on a local feature consistency optimization algorithm.

[0025] As an optional implementation, the data fusion of the alignment results of each level by the grid interpolation method to generate a unified feature distribution model includes:

[0026] Determining a fusion weight when performing the data fusion based on the distribution characteristics of the surface morphology data and the ultrasonic detection data;

[0027] Based on the fusion weight, the alignment results of each level are divided into a plurality of interpolation regions; the region types of the interpolation regions include: data-intensive regions and data-sparse regions;

[0028] Interpolation processing is performed within each of the interpolation areas.

[0029] As an optional implementation manner, determining the fusion weight includes:

[0030] Determining the weight of the surface topography data based on the surface defect distribution density of the target metal material;

[0031] The weight of the ultrasonic detection data is determined based on the depth and signal strength of the internal defects of the target metal material.

[0032] As an optional implementation, the interpolation process includes:

[0033] For the data-intensive area, high-order polynomial interpolation is adopted;

[0034] For the data sparse area, a distance weighted interpolation method or a low-order interpolation method is adopted.

[0035] As an optional implementation manner, the extracting key defect information of the target metal material further includes:

[0036] Based on the defect location data, generating input data for edge detection; wherein the input data for edge detection includes: three-dimensional coordinates, contour features, and depth information of the defect;

[0037] Using the Canny edge detection algorithm to process the defect area in the input data, and extract the defect edge information of the target metal material;

[0038] Based on the defect edge information, key defect information of the target metal material is extracted.

[0039] As an optional implementation, before extracting the key defect information of the target metal material based on the defect edge information, the method further includes: optimizing the edge detection result based on the three-dimensional coordinate constraint in the defect location data; wherein the optimizing the edge detection result includes:

[0040] Identify and remove isolated edge points by calculating the neighborhood consistency score for each edge point;

[0041] The morphological dilation algorithm is used to connect the discontinuous edge parts in the edge detection results to generate continuous edge information.

[0042] According to the three-dimensional coordinate constraints in the defect location data, the spatial range of the edge points is defined, and the edge points that deviate from the range are adjusted.

[0043] In a second aspect, the present invention provides a metal material defect identification and analysis system based on multiple detection technology, comprising:

[0044] A first acquisition module is used to detect the surface morphology of the target metal material using a preset microscope device to obtain surface morphology data; wherein the surface morphology data includes: image data of the surface of the target metal material, and size, shape and distribution information of surface defects of the target metal material;

[0045] A second acquisition module is used to detect the internal structure of the target metal material based on ultrasonic detection technology to generate ultrasonic detection data; wherein the ultrasonic detection data includes: waveform amplitude, frequency characteristics, depth and shape information of the internal defects of the target metal material;

[0046] The first processing module is used to perform feature alignment and fusion processing on the surface morphology data and the ultrasonic detection data based on the multi-grid method to generate solution data of ultrasonic wave field propagation; based on the solution data, determine the position, shape and size information of the potential defect, and generate defect location data including three-dimensional coordinates and contour features;

[0047] A second processing module is used to extract key defect information of the target metal material based on the defect location data using an edge detection algorithm; wherein the key defect information includes the contour and depth information of the defect;

[0048] A generation module is used to determine the defect type of the target metal material based on the key defect information and generate detection result information corresponding to the target metal material.

[0049] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention integrates microscope detection and ultrasonic detection technology to achieve comprehensive detection and accurate identification of surface and internal defects. Microscope detection obtains surface morphology data, and ultrasonic detection generates internal structure data. The combination of the two greatly improves the comprehensiveness and accuracy of detection. The multi-grid method is used to fuse data from different sources to generate a unified feature distribution model to ensure data consistency and integrity. Based on the fusion model, a mathematical solution model for ultrasonic wave field propagation is established to accurately locate the position, shape and size information of the defect. At the same time, through feature matching and classification algorithms, the defect type can be quickly and accurately determined. This method is suitable for complex scenarios in multiple industries, and solves the problems of insufficient detection accuracy, difficult data fusion and low classification efficiency in the prior art, and significantly improves the applicability and reliability of detection technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A flowchart of a metal material defect identification and analysis method based on multiple detection technologies provided by an embodiment of the present invention;

[0051] Figure 2 A flow chart of a method for generating solution data for ultrasonic wave field propagation provided for the implementation of the present invention;

[0052] Figure 3 Schematic diagram of a metal material defect recognition and analysis system based on multiple detection technologies provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0054] See also Figure 1 , Figure 1 A flowchart of a metal material defect recognition and analysis method based on multiple detection technologies provided by an embodiment of the present invention. The present invention provides a metal material defect recognition and analysis method based on multiple detection technologies. The method comprehensively utilizes microscope detection and ultrasonic detection technology to comprehensively analyze the surface and internal defects of the target metal material, and realizes accurate positioning and classification of defects through data fusion and algorithm processing. The method includes steps S101 to S105, specifically:

[0055] S101: Using a preset microscope device to detect the surface morphology of a target metal material to obtain surface morphology data; wherein the surface morphology data includes: image data of the surface of the target metal material, and size, shape and distribution information of surface defects of the target metal material;

[0056] S102: Detecting the internal structure of the target metal material based on ultrasonic detection technology to generate ultrasonic detection data; wherein the ultrasonic detection data includes: waveform amplitude, frequency characteristics, depth and shape information of the internal defects of the target metal material;

[0057] S103: performing feature alignment and fusion processing on the surface morphology data and the ultrasonic detection data based on a multi-grid method to generate solution data for ultrasonic wave field propagation; based on the solution data, determining the position, shape and size information of the potential defect, and generating defect location data including three-dimensional coordinates and contour features;

[0058] S104: extracting key defect information of the target metal material based on the defect location data using an edge detection algorithm; wherein the key defect information includes defect profile and depth information;

[0059] S105: Based on the key defect information, determine the defect type of the target metal material, and generate detection result information corresponding to the target metal material.

[0060] Regarding S101 above:

[0061] In a specific implementation, the present invention uses a preset microscope device to detect the surface morphology of the target metal material and obtain surface morphology data. The microscope device may include multiple types such as an optical microscope, a scanning electron microscope (SEM) or an atomic force microscope (AFM).

[0062] The specific choice can be determined according to the surface characteristics of the target metal material and the detection requirements. For example, optical microscopes are suitable for preliminary observation of surface features over a large range, SEM is suitable for analyzing high-resolution details, and AFM can provide nanoscale surface characteristic data. Before testing, the metal material needs to be cleaned to ensure that its surface is not disturbed by contaminants. For samples with higher roughness, mechanical polishing or chemical polishing can also be used to improve the surface condition.

[0063] Among them, during the detection process, the high-resolution imaging function of the microscope device can be used to gradually scan the surface of the target metal material to obtain image data containing complete surface morphology information. Through the built-in or external data processing system, the collected image is analyzed using image processing algorithms (such as edge detection and image segmentation) to extract the key morphological features of the metal material surface. The morphological features include the size, shape and distribution information of the defects. For example, the linear size (length and width) and area of ​​the defect can be measured by the ruler calibration method; the shape of the defect (such as cracks, pores or scratches, etc.) can be identified by the geometric fitting algorithm; the distribution density and spatial distribution law of the defect on the material surface can be determined by statistical analysis.

[0064] In specific implementation, the combined use of multiple microscopes can effectively improve the detection accuracy and efficiency. For example, based on the large-scale preliminary scanning of the optical microscope, the SEM is used to perform high-resolution imaging of key areas to further analyze the details of tiny defects; at the same time, the AFM can be combined to perform in-depth analysis of nano-level defects, thereby achieving multi-scale detection of surface morphology from macro to micro. This method can not only improve the comprehensiveness and accuracy of detection, but also provide a data basis for subsequent internal defect detection and data fusion processing.

[0065] In this way, the surface morphology data of the target metal material can be accurately acquired, ensuring the complete characteristic information of surface defects, and providing a reliable basis for subsequent defect analysis and classification.

[0066] Regarding S102 above:

[0067] In a specific implementation, the present invention detects the internal structure of the target metal material based on ultrasonic testing technology to obtain ultrasonic testing data. Among them, ultrasonic testing technology is a non-destructive testing (NDT) method that analyzes the structural integrity and potential defects inside the material through the propagation characteristics of high-frequency sound waves in metal materials. Ultrasonic testing data includes waveform amplitude, frequency characteristics, depth and shape information of internal defects of the target metal material.

[0068] In the specific implementation, it is first necessary to select appropriate ultrasonic testing equipment, such as a portable ultrasonic detector or an industrial-grade ultrasonic imaging device. According to the characteristics of the target metal material, the operating frequency, waveform type and probe type of the ultrasonic detection are set. Common probe types include straight probes, oblique probes and focused probes. Among them, the straight probe is suitable for internal detection of relatively flat materials, the oblique probe is suitable for evaluating defects in welds, and the focused probe is suitable for accurate detection of small volume areas.

[0069] During the detection process, the ultrasonic detector emits high-frequency sound waves, which enter the target metal material through the probe. When the sound waves propagate to the defect location, due to the discontinuity of the acoustic impedance, partial reflection or scattered waves will be generated. The received echo signal is processed by the instrument to generate detection data. The characteristics of the echo signal (such as waveform amplitude, frequency distribution, and time delay) reflect the characteristics of the internal defects of the material.

[0070] For example, changes in waveform amplitude can reflect the size and nature of the defect, the characteristics of the frequency distribution can be used to determine the material of the defect, and time delay can determine the depth information of the defect.

[0071] In order to improve the detection accuracy and reliability, a variety of signal processing algorithms can be used to optimize the ultrasonic detection data. For example, by performing spectrum analysis on the echo signal through fast Fourier transform (FFT), the frequency characteristics of the signal can be extracted; by analyzing the time-amplitude curve, the shape and depth information of the defect can be extracted. In order to eliminate noise interference, the signal can be denoised by combining filtering algorithms (such as bandpass filtering and low-pass filtering) to enhance the significance of the defect.

[0072] In addition, in order to ensure the accuracy and comprehensiveness of the data, a multi-angle and multi-directional ultrasonic testing strategy can be adopted. Specifically, the probe is placed at different positions of the target metal material and sound waves are emitted at multiple angles to capture multi-directional echo signals from defects. By integrating multi-directional detection data, the accurate judgment of defect location and shape information can be improved.

[0073] For example, taking a carbon steel plate with a thickness of 20 mm as an example, ultrasonic testing technology is used to detect possible crack defects inside it. First, select appropriate ultrasonic testing equipment. For example, an industrial-grade ultrasonic detector is selected, equipped with a straight probe with a frequency of 5 MHz, a probe diameter of 10 mm, and a longitudinal wave mode. The operating frequency is set to 5 MHz to take into account both the detection depth and resolution. The pulse echo method is used to emit short pulse ultrasonic waves. In order to ensure that the ultrasonic wave is effectively transmitted into the material, a water-soluble coupling agent is applied between the probe and the surface of the steel plate.

[0074] During the inspection process, the probe is placed vertically on the surface of the steel plate and moved along the preset scanning path. The scanning area is 100 mm × 100 mm. The linear scanning method is adopted, and the probe is moved every 5 mm to cover the entire inspection area. In terms of equipment settings, the pulse repetition frequency is set to 1000 Hz and the gain is set to 40 dB to ensure moderate signal strength. The time window is set to 0 to 15 microseconds, and the corresponding detection depth range is 0 to 30 mm.

[0075] During the data acquisition process, the waveform of the echo signal, including the amplitude and arrival time, is recorded at each scanning point. In order to improve the signal quality, the echo signal is processed. First, a bandpass filter is used to filter out noise signals below 1 MHz and above 6 MHz; then, the echo signal is subjected to a fast Fourier transform (FFT) to obtain the spectrum information. In terms of defect identification, a threshold for the echo amplitude is set. When the echo amplitude exceeds 50% of the full amplitude, it is determined that a defect may exist. The depth of the defect is calculated based on the echo arrival time.

[0076] For example, in the test results, a signal with a significantly increased echo amplitude was detected at the coordinates (50 mm, 50 mm). The defect depth was calculated to be about 10 mm. According to the echo amplitude and signal expansion, the lateral size of the defect was estimated to be about 5 mm. Through spectrum analysis, it was found that the spectrum characteristics of the defect were consistent with cracks, and it was inferred to be an internal crack.

[0077] In order to verify the test results, the probe angle was adjusted and the oblique probe was used to test again at a 35° incident angle to confirm the existence and location of the defect. Integrating the data from multi-angle tests can provide complete information on the shape and size of the defect.

[0078] In this way, ultrasonic testing can generate a complete data set including waveform amplitude, frequency characteristics, depth and shape information of internal defects of target metal materials. Compared with traditional single-direction ultrasonic testing, the ultrasonic testing method of the present invention can more comprehensively reflect the internal structural characteristics and defect information of metal materials.

[0079] Regarding S103 above:

[0080] The present invention performs feature alignment and fusion processing on surface morphology data and ultrasonic detection data based on the multi-grid method, generates solution data of ultrasonic wave field propagation, and determines the position, shape and size information of potential defects based on the solution data, thereby generating defect location data containing three-dimensional coordinates and contour features.

[0081] In the specific implementation, the surface morphology data and ultrasonic detection data obtained in S101 and S102 are first imported into the feature fusion system. The surface morphology data includes the image information of the metal material surface and the geometric characteristics of the surface defects (such as size, shape and distribution information), while the ultrasonic detection data provides the depth, shape, waveform amplitude and frequency characteristic information of the internal defects of the material.

[0082] Among them, these data often come from different detection technologies and have different resolutions and data formats, so they need to be aligned and fused.

[0083] See also Figure 2 , Figure 2 A flow chart of a method for generating solution data for ultrasonic wave field propagation provided for the implementation of the present invention, as an optional implementation, the method for generating solution data for ultrasonic wave field propagation includes the following steps S201 to S204, wherein:

[0084] S201: performing layered processing on the surface morphology data and the ultrasonic detection data respectively to extract feature information at different scales;

[0085] S202: aligning the features of the surface topography data and the ultrasonic detection data at each level by a multi-grid method, wherein the alignment process includes feature point matching and geometric transformation;

[0086] S203: After completing the alignment of each level, data fusion is performed on the alignment results of each level through a grid interpolation method to generate a unified feature distribution model;

[0087] S204: Establishing a mathematical solution model for ultrasonic wave field propagation according to the characteristic distribution model, and generating solution data for ultrasonic wave field propagation.

[0088] As an optional implementation, the multi-grid method also includes: extracting significant feature points and generating a surface morphology feature point set based on the geometric constraints of the surface morphology of the target metal material; generating an internal structure feature point set based on the depth information and waveform amplitude characteristics of the internal structure in the ultrasonic detection data; and performing multi-scale morphological matching between the surface morphology feature point set and the internal structure feature point set through a morphological constraint registration method.

[0089] In specific implementation, the operation steps of the multi-grid method include feature point extraction, feature point set generation, and multi-scale matching under morphological constraints. First, extract significant feature points from the surface morphology data of the target metal material. Surface morphology data is usually represented in the form of a three-dimensional point cloud, where each point contains spatial coordinates and local curvature information. The extraction criteria for significant feature points are defined by geometric constraints. For example, when the curvature change rate of a point is greater than a set threshold (such as 0.02), the point is marked as a significant feature point. The curvature analysis algorithm is used to calculate the entire point cloud data point by point to extract the point set with significant curvature changes. The output result is a surface morphology feature point set in the format of three-dimensional coordinate data (X, Y, Z), with the local curvature value of each point attached.

[0090] Next, extract the internal structure feature point set from the ultrasonic detection data. Ultrasonic detection data usually contains the amplitude and depth information of the reflected waveform. By detecting the peak value of the reflected signal, the significant feature points of the ultrasonic wave in the defect area inside the material are identified. Specifically, the original signal is first subjected to noise filtering and signal smoothing, and the moving average filter method is used to eliminate high-frequency noise; then, by setting the threshold of the reflected wave amplitude (such as the signal amplitude is greater than 70% of the maximum amplitude), the significant peak points in the ultrasonic signal are extracted. Finally, combined with the corresponding depth data, the internal structure feature point set is generated, and the output results include three-dimensional coordinates (X, Y, Z) and the reflection intensity value of each point.

[0091] Then, the morphological constraint registration method is used to perform multi-scale morphological matching on the surface morphology feature point set and the internal structure feature point set. First, according to the spatial distribution characteristics of the point set, it is divided into macro-level and micro-level feature point sets. The feature points at the macro level are mainly used for global alignment. The rigid body transformation model is used to achieve preliminary registration. The point cloud alignment results are optimized by the least squares method, and the error range is controlled within 0.05 mm. For the feature points at the micro level, a non-rigid body transformation model is used, combined with morphological constraints (such as the distance and angle relationship between feature points), to perform fine matching on the local area. The iterative process of the registration algorithm is terminated when the error converges to less than 0.01 mm.

[0092] Taking a certain alloy material as an example, its surface morphology data is acquired by a laser scanner, and the generated point cloud contains about 5 million points; the ultrasonic detection data is collected by an ultrasonic instrument, and the depth resolution of the generated reflection waveform is 0.1 mm. In actual operation, about 100,000 significant feature points are extracted for registration. Through the above method, the surface morphology feature point set and the internal structure feature point set are successfully aligned, and the generated matching results include complete three-dimensional point cloud data (X, Y, Z) and corresponding attribute data (such as curvature value, reflection intensity value).

[0093] Through the above steps, the present invention realizes high-precision alignment and fusion of surface and internal features, providing a reliable basis for subsequent data processing and defect analysis. The method has significant adaptability and operability, and is suitable for different types of metal materials and a variety of complex detection environments.

[0094] As an optional implementation, multi-scale morphological matching includes: using a surface fitting algorithm to fit a set of surface morphology feature points into a continuous surface to generate a constrained surface model; based on the constrained surface model, using a deep feature optimization algorithm to transform and adjust the internal structure feature point set to achieve global feature alignment; in response to the completion of global feature alignment, dynamically adjusting the surface morphology features and ultrasonic detection features based on a local feature consistency optimization algorithm.

[0095] In specific implementation, the method of multi-scale morphological matching includes three main steps: surface fitting, global feature alignment, and local feature consistency optimization. First, the surface morphology feature point set is fitted into a continuous surface using a surface fitting algorithm to generate a constrained surface model. The surface morphology feature point set comes from the above-mentioned feature point extraction process and is in the form of three-dimensional point cloud data. During the fitting process, the B-spline surface fitting algorithm is used to calculate the fitting surface of the point cloud data using the least squares method as the optimization criterion. The input of the algorithm includes point cloud coordinates and neighborhood curvature information, and the fitting error is controlled within 0.05 mm. After the fitting is completed, the generated constrained surface model is expressed in the form of a B-spline equation and further converted into a three-dimensional mesh model for subsequent processing.

[0096] Next, based on the constrained surface model, the deep feature optimization algorithm is used to transform and adjust the internal structure feature point set to achieve global feature alignment. First, the initial matching error between the internal structure feature point set and the constrained surface model is calculated, and the shortest distance from the point to the surface is used as the error measurement standard. Subsequently, the iterative closest point (ICP) algorithm is used for optimization. Each iteration calculates the projection point of the feature point to the constrained surface and updates the coordinate position of the feature point set. To ensure the accuracy of global alignment, the error convergence condition is set, and the optimization is terminated when the average projection error of all feature points is less than 0.01 mm. The final global alignment result includes the coordinate transformation matrix of the feature point set and its aligned three-dimensional coordinates.

[0097] After completing the global alignment, the surface morphology features and ultrasonic detection features are further dynamically adjusted based on the local feature consistency optimization algorithm. Local optimization is based on the neighborhood consistency of each feature point. The local adjustment strategy is determined by calculating the distance change and direction vector difference of the feature points in the neighborhood. Specifically, for feature points with large deviations (deviations greater than 0.02 mm), higher adjustment weights are assigned; for feature points with small deviations, their original coordinates are maintained. A weighted dynamic adjustment algorithm is used in the optimization process to gradually reduce the differences between local feature points, so that the optimization results have higher consistency in the global range.

[0098] For example, taking a certain aluminum alloy material as an example, its surface morphology feature point set is obtained by a laser scanner, which contains about 300,000 three-dimensional points; the internal structure feature point set is collected by an ultrasonic detector, which contains 100,000 depth feature points. Through the above method, the constrained surface model first generated is represented by the B-spline fitting result, and the global alignment of the internal feature point set is achieved through the ICP algorithm; finally, the average point-to-point error is reduced from 0.03 mm to 0.008 mm through the local feature consistency optimization algorithm. The optimized feature point set can accurately describe the geometric morphology of the material surface and the spatial distribution of internal defects, providing an accurate basis for subsequent data fusion and defect analysis.

[0099] As an optional implementation, data fusion is performed on the alignment results of each level through a grid interpolation method to generate a unified feature distribution model including:

[0100] Determining a fusion weight when performing the data fusion based on the distribution characteristics of the surface morphology data and the ultrasonic detection data;

[0101] Based on the fusion weight, the alignment results of each level are divided into a plurality of interpolation regions; the region types of the interpolation regions include: data-intensive regions and data-sparse regions;

[0102] Interpolation processing is performed within each of the interpolation areas.

[0103] As an optional implementation, determining the fusion weight includes:

[0104] Determining the weight of the surface topography data based on the surface defect distribution density of the target metal material;

[0105] The weight of the ultrasonic detection data is determined based on the depth and signal strength of the internal defects of the target metal material.

[0106] As an optional implementation, the interpolation process includes:

[0107] For the data-intensive area, high-order polynomial interpolation is adopted;

[0108] For the data sparse area, a distance weighted interpolation method or a low-order interpolation method is adopted.

[0109] In a specific implementation, the present invention uses a grid interpolation method to perform data fusion on the alignment results of each level to generate a unified feature distribution model. First, based on the distribution characteristics of the surface morphology data and the ultrasonic detection data, the fusion weight during data fusion is determined. For the surface morphology data, the weight value is determined by calculating the distribution density of surface defects of the target metal material. Specifically, the surface morphology data is divided into multiple unit areas according to a spatial grid, and the number density of defect points in each area is counted. For areas with higher density (such as more than 50 feature points per unit area), a higher fusion weight (such as 0.7-1) is assigned; for areas with lower density (such as less than 20 feature points), the weight value is set to 0.3-0.5.

[0110] At the same time, for ultrasonic testing data, the weight value is calculated according to the depth and signal strength of the internal defect. Defects with greater depth usually reflect more serious internal damage, so they are given a higher weight (such as 0.6-0.8); the stronger the signal strength of the defect point, the more significant the reflection characteristics, and the weight value is increased accordingly (such as when the signal strength exceeds 80% of the standard value, the weight is set to 0.8-1). Through normalization processing, the weight distribution range of the surface morphology data and the ultrasonic testing data is adjusted to be consistent.

[0111] Next, the interpolation area is divided according to the fusion weight. Specifically, the spatial distribution of the alignment results is divided into data-intensive areas and data-sparse areas. Data-intensive areas are defined as areas with more than 30 feature points per unit grid; data-sparse areas are areas with less than 10 feature points. For data-intensive areas, high-order polynomial interpolation methods are used, such as using cubic spline interpolation to fit the feature distribution in the area; for data-sparse areas, distance-weighted interpolation methods are used to estimate feature values ​​based on the weight distribution of surrounding points.

[0112] After performing interpolation processing in each interpolation area, a unified feature distribution model is generated. This model combines the spatial characteristics of surface morphology data and ultrasonic testing data, and can accurately reflect the surface and internal defect distribution characteristics of the target metal material. The output of the model is represented in the form of a three-dimensional grid, and the node data includes the fused feature values ​​and their spatial coordinates.

[0113] Taking a certain stainless steel material as an example, its surface morphology data is collected by a laser scanner, and the ultrasonic detection data is obtained by an ultrasonic reflectometer. Through the above method, the fused feature distribution model achieves high-precision interpolation processing in the key defect areas on the surface and inside. The results show that the matching degree between the surface morphology and the internal defect information is improved by 30%, and the eigenvalue estimation error in the data sparse area is controlled within 0.05 mm. The generated feature distribution model provides a basis for subsequent defect analysis and classification.

[0114] Regarding the above S204:

[0115] In the specific implementation, it is necessary to establish a mathematical solution model for ultrasonic wave field propagation from the fused feature distribution model. The feature distribution model is represented in the form of a three-dimensional grid, and the node data includes the spatial coordinates of each node, material properties (such as density, elastic modulus, Poisson's ratio) and defect feature information (such as defect type, size, shape and depth).

[0116] First, mesh the feature distribution model. Select a three-dimensional unit type suitable for the wave problem (such as tetrahedral units or hexahedral units) to accommodate complex geometric shapes and material property changes. In defect areas and complex morphology areas, use finer meshing to improve calculation accuracy; in defect-free uniform areas, use coarser meshes to improve calculation efficiency. The meshing should meet the numerical accuracy requirements of the wave problem, that is, the unit size should be small enough to capture the wavelength characteristics of the ultrasound.

[0117] Then, define the material properties. In the defect-free area, set the material parameters such as density, elastic modulus, and Poisson's ratio to reflect the actual physical properties of the material. In the defect area, adjust the material properties according to the defect type and characteristics. For example, for the crack area, the elastic modulus of the material can be set to close to zero, indicating that the area is a void or damage; for the inclusion area, different densities and elastic moduli can be set to simulate the discontinuity and heterogeneity of the material.

[0118] Next, set the boundary conditions. Specify the location and form of ultrasonic excitation on the material surface. The excitation can be displacement excitation or force excitation. The excitation signal can be Gaussian pulse, sine wave or other suitable waveform. The frequency is set according to the detection requirements and material properties. Set absorbing boundary conditions (such as perfect matching layers) at the boundaries of the model to prevent boundary reflections from affecting the calculation results. The initial conditions are usually set to the material being at rest, with zero displacement and velocity.

[0119] The control equation for the propagation of ultrasonic wave field is established. The propagation of ultrasonic wave in elastic medium satisfies the elastic wave equation:

[0120]

[0121] in, is the density of the material, is the displacement vector, is the stress tensor, and the strain Connected by constitutive relations: , is the elastic stiffness tensor, is the volume force density, is the divergence of the stress tensor.

[0122] In the numerical solution process, time and space are discretized. Time discretization can use explicit or implicit time integration methods, and the time step must meet stability conditions and accuracy requirements. Space discretization uses the finite element method to assemble the mass matrix and stiffness matrix. At each time step, the displacement field, velocity field, and acceleration field are solved.

[0123] For example, in the numerical solution process, the explicit central difference method or implicit Newmark method is used to discretize time. It should be noted that the time step must meet the stability condition (CFL condition). The finite element method is used to discretize space and assemble the overall mass matrix and stiffness matrix. At each time step, the displacement field is solved , update velocity and acceleration.

[0124] It should be noted that in the process of numerical solution, the actual physical properties of the material, such as anisotropy and damping effect, need to be considered to improve the accuracy of the model.

[0125] In this way, through numerical solution, the propagation process of ultrasonic waves inside the material is obtained, including data such as displacement field, stress field and velocity field at different time steps. These solution data can be used for wave field visualization, defect identification and signal processing. Through wave field visualization, the propagation path, reflection and scattering of ultrasonic waves in the material can be displayed, and the impact of defects on the wave field can be intuitively observed. In terms of defect identification, the abnormal changes of the wave field in the defect area are analyzed, such as enhanced reflection, weakened transmission, increased scattering, etc., to further confirm the location and nature of the defect. In terms of signal processing, time series signals are extracted at the receiving point (such as certain positions on the surface of the material) for spectrum analysis and time domain analysis.

[0126] It should be noted that in order to ensure the accuracy and efficiency of calculation, the grid division, material properties, boundary conditions and numerical solution parameters should be reasonably set according to the specific material characteristics and detection requirements. For models with larger calculation scales, high-performance computing platforms or parallel computing technologies can be used to meet the needs of computing resources.

[0127] For example, in practical applications, sensitivity analysis is performed on key parameters (such as mesh size, time step, material properties, etc.) to ensure the stability of the model and the reliability of the results.

[0128] For example, in defective areas and complex morphology areas, a finer grid is used, and the unit size is 1 / 10 to 1 / 15 of the wavelength; in defect-free uniform areas, a coarser grid is used. The number of units after the entire model is divided is about 1 million to 5 million to ensure calculation accuracy.

[0129] Through the above method, a mathematical solution model for ultrasonic wave field propagation is established from the characteristic distribution model, and solution data for ultrasonic wave field propagation is generated, thereby achieving accurate simulation and analysis of internal defects of the target metal material.

[0130] In a specific implementation, the present invention further determines the location, shape and size information of potential defects based on the solution data of ultrasonic wave field propagation, and generates defect location data containing three-dimensional coordinates and contour features. First, wave field analysis is performed using the solution data. The solution data contains the temporal and spatial distribution information of ultrasonic wave propagation inside the metal material. By analyzing the time series and spatial distribution of the wave field, the areas with abnormal changes in the wave field can be identified, which usually correspond to defects inside the material.

[0131] In the specific implementation, the wave field data processing method is used to conduct in-depth analysis of the solution data in the time domain and frequency domain. In the time domain analysis, abnormal changes in the waveform, such as the increase of reflected waves or the decrease of transmitted waves, are observed to identify potential defect signals. In the frequency domain analysis, the time series is Fourier transformed to obtain spectrum information and identify abnormal changes in frequency components, which helps to distinguish different types of defects.

[0132] Subsequently, the defects are identified and located. The location of the abnormal area is determined by comparing the changes in the wave field data at different locations. The three-dimensional coordinates of the defect in the material are calculated using the propagation speed of the ultrasonic wave in the material and the signal arrival time difference to achieve precise positioning. Next, the shape and size of the defect are estimated. In the identified defect area, image processing algorithms such as edge detection and regional growth are used to extract the spatial contour of the defect. Based on the obtained contour data, the length, width, depth and other dimensional parameters of the defect are calculated to fully describe the geometric characteristics of the defect.

[0133] When generating defect location data, the defect information is organized into a structured data form, including three-dimensional coordinates, contour features, and size information. Specifically, three-dimensional coordinates refer to the X, Y, and Z coordinates of the center of the defect; contour features can be represented in the form of point clouds, grids, or mathematical surfaces, describing the spatial form of the defect in detail; and size information includes key parameters such as the length, width, and depth of the defect. These data are stored in a database or file for subsequent analysis and processing.

[0134] Taking the detection of internal cracks in a 20 mm thick carbon steel plate as an example, the application of the above method is specifically explained. Through wave field analysis, the wave field is found to have abnormal reflections at X = 50 mm, Y = 75 mm, and Z = 10 mm using the solution data. By calculating the arrival time of the reflected wave and the known propagation speed of the ultrasonic wave in the carbon steel plate, the three-dimensional coordinates of the crack are determined to be (50 mm, 75 mm, 10 mm). In the shape and size estimation, the edge detection algorithm is used to extract the contour of the crack, and it is found that the crack is linear, with a length of about 30 mm and a width of about 2 mm. Subsequently, the above information is organized into defect location data, including three-dimensional coordinates (X = 50 mm, Y = 75 mm, Z = 10 mm), contour features (spatial curve data of the crack), and size information (length 30 mm, width 2 mm).

[0135] In order to verify the accuracy of the results, the generated defect location data was compared with the actual detection results, and it was found that the positioning error was within the allowable range, proving the effectiveness of the method. Through the above process, the location, shape and size information of the potential defects were successfully determined based on the solution data of the ultrasonic wave field propagation, and defect location data containing three-dimensional coordinates and contour features were generated, providing reliable data support for subsequent defect analysis and material evaluation.

[0136] Regarding S104 above:

[0137] The present invention uses edge detection algorithm to extract key defect information of target metal materials based on defect location data. The key defect information includes the contour and depth information of the defect. By accurately extracting key features, it provides a data basis for subsequent defect classification and detection result generation. Defect location data is used as input, including the three-dimensional coordinates and contour feature information of potential defects. These data are derived from the defect location data generated by S103 above. The surface morphology data and ultrasonic detection data have been aligned and fused by the multi-grid method, which can reflect the spatial distribution characteristics of defects.

[0138] In the specific implementation, the defect location data includes three-dimensional coordinates and contour feature information, which reflects the spatial distribution and appearance characteristics of the defect. First, based on the defect location data, the defect information is projected onto a two-dimensional plane for edge detection, while retaining the depth information for subsequent processing. The projected data includes:

[0139] The contour features of the defect after the three-dimensional coordinates are mapped to the two-dimensional image plane;

[0140] The depth information of the defect is stored as an additional attribute.

[0141] As an optional implementation manner, the extracting key defect information of the target metal material further includes:

[0142] Based on the defect location data, generating input data for edge detection; wherein the input data for edge detection includes: three-dimensional coordinates, contour features, and depth information of the defect;

[0143] Using the Canny edge detection algorithm to process the defect area in the input data, and extract the defect edge information of the target metal material;

[0144] Based on the defect edge information, key defect information of the target metal material is extracted.

[0145] As an optional implementation, before extracting the key defect information of the target metal material based on the defect edge information, the method further includes: optimizing the edge detection result based on the three-dimensional coordinate constraint in the defect location data; wherein the optimizing the edge detection result includes:

[0146] Identify and remove isolated edge points by calculating the neighborhood consistency score for each edge point;

[0147] The morphological dilation algorithm is used to connect the discontinuous edge parts in the edge detection results to generate continuous edge information.

[0148] According to the three-dimensional coordinate constraints in the defect location data, the spatial range of the edge points is defined, and the edge points that deviate from the range are adjusted.

[0149] In order to improve the accuracy of edge detection results before extracting key defect information of the target metal material based on defect edge information, the edge detection results can be optimized based on the three-dimensional coordinate constraints in the defect location data.

[0150] In the specific implementation, the defect location data and the initial edge detection results can be obtained first. The defect location data comes from the previous step, including the three-dimensional coordinate information (X, Y, Z) of the defect area, which clarifies the position and range of the defect in space; the edge detection results are obtained through the Canny edge detection algorithm, including the two-dimensional pixel coordinates (u, v) of the edge point.

[0151] In order to eliminate noise and isolated points, it is necessary to calculate the neighborhood consistency score for each edge point to identify and remove isolated edge points. Exemplarily, with each edge point as the center, a 5×5 pixel window is selected as its neighborhood, and the number of other edge points n (excluding the center point) in the window is counted, and the number is used as the neighborhood consistency score S, that is, S = n. Set the scoring threshold T = 2. When S is less than the scoring threshold, the edge point is judged as an isolated point and removed from the edge detection result. Through this step, isolated edge points caused by noise are effectively identified and removed, and the reliability of edge data is improved.

[0152] Next, for the edge discontinuity in the edge detection results, the morphological dilation algorithm is used to connect the discontinuous edge parts to generate continuous edge information. A 3×3 cross-shaped structural element (center is 1, horizontal and vertical directions are 1, and diagonal is 0) is selected to dilate the edge image to expand the edge range so that adjacent edge segments can be connected. In order to restore the fineness of the edge, a morphological thinning or corrosion operation is then performed to restore the edge to a single pixel width. Through the combination of dilation and thinning, the broken edge segments are successfully connected, continuous edge information is generated, and the integrity of the edge is improved.

[0153] Finally, according to the three-dimensional coordinate constraints in the defect location data, the spatial range of the edge points is defined, and the edge points that deviate from the range are adjusted. First, based on the defect location data, the minimum bounding box (BoundingBox) of the defect area is determined to obtain the range in the X, Y, and Z directions ( , ; , ; , Then, the 2D coordinates (u, v) of the edge points are mapped back to the 3D space and back-projected using the camera model or sensor parameters to obtain the corresponding 3D coordinates (X, Y, Z). For edge points that are beyond the predefined spatial range, their coordinates are adjusted to the nearest bounding box boundary. For example, if the X coordinate exceeds , then set it to For edge points that deviate beyond the maximum allowable deviation (eg, 2 mm), they are removed from the edge data.

[0154] This spatial constraint processing ensures that the edge points are accurately located within the actual spatial range of the defect, avoiding edge point deviation due to noise or false detection.

[0155] After the above optimization processing, the edge detection results more accurately reflect the true shape of the defects, providing a data basis for the subsequent extraction of key defect information and evaluation of the impact of defects on material properties.

[0156] For example, when detecting cracks inside a steel plate, the above method improves the accuracy of edge detection results by about 20%, and key parameters such as crack length, width and depth can be accurately measured, providing reliable data support for the safety assessment and subsequent processing of the steel plate.

[0157] Regarding the above S105:

[0158] In the present invention, the defect type of the target metal material is judged based on the key defect information, and the corresponding detection result information is generated. The key defect information comes from the processing result of S104, including the three-dimensional profile, depth range and local characteristics of the defect. Through comprehensive analysis of this information, accurate classification of the defect type of the metal material can be achieved.

[0159] In specific implementation, the defect feature matching method can be used to compare the extracted key defect information with the preset defect type database. The defect type database pre-stores standard feature parameters of various common defects (such as cracks, pores, inclusions, etc.), including size range, shape characteristics and depth distribution law.

[0160] For example, cracks usually appear linear or branched, pores are mostly round or elliptical, and inclusions have irregular shapes. By comparing the actual characteristics of the defect with the matching degree of the standard characteristic parameters in the database, the type of defect can be preliminarily determined.

[0161] In addition, based on feature matching, the present invention further uses a classification algorithm to refine the defect type. Specifically, the comprehensive features of the defect (such as three-dimensional profile parameters, depth range, local edge characteristics, etc.) are used as input, and machine learning algorithms such as support vector machines (SVM) or decision trees are used for classification. This classification algorithm can quickly and accurately distinguish different types of defects through training and optimization. For example, based on the depth range and shape parameters of the defect, it can be determined whether it is a surface defect (such as a crack) or an internal defect (such as a pore or inclusion).

[0162] Once the defect type classification is completed, the system generates the test result information of the target metal material. The test result information includes, for example:

[0163] Defect Type: Clearly describe the category of defects (such as cracks, pores, inclusions, etc.).

[0164] Defect characteristics: including the three-dimensional position, size, depth range and shape description of the defect.

[0165] Quality assessment: Combine the distribution density, size and location of defects to evaluate whether the target metal material meets the preset engineering use standards. For example, for materials with a specific crack length or pore density exceeding the threshold, the test results will mark it as non-compliant.

[0166] In this way, the present invention can realize efficient classification and comprehensive detection of metal material defects, and provide a reliable basis for material quality assessment. This method not only improves the accuracy of the test results, but also has the ability to quickly handle complex defects, and is particularly suitable for large-scale material testing in industrial environments.

[0167] Based on the same inventive concept, the embodiment of the present invention also provides a metal material defect identification and analysis system based on multiple detection technologies corresponding to the metal material defect identification and analysis method based on multiple detection technologies. Since the principle of solving the problem by the system in the embodiment of the present invention is similar to the above-mentioned metal material defect identification and analysis method based on multiple detection technologies, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be repeated.

[0168] Reference Figure 3 As shown, Figure 3 A schematic diagram of a metal material defect recognition and analysis system based on multiple detection technologies provided by an embodiment of the present invention, the system comprising:

[0169] The first acquisition module 10 is used to detect the surface morphology of the target metal material using a preset microscope device to obtain surface morphology data; wherein the surface morphology data includes: image data of the surface of the target metal material, and size, shape and distribution information of surface defects of the target metal material;

[0170] The second acquisition module 20 is used to detect the internal structure of the target metal material based on ultrasonic detection technology to generate ultrasonic detection data; wherein the ultrasonic detection data includes: waveform amplitude, frequency characteristics, depth and shape information of the internal defects of the target metal material;

[0171] The first processing module 30 is used to perform feature alignment and fusion processing on the surface morphology data and the ultrasonic detection data based on the multi-grid method to generate solution data of ultrasonic wave field propagation; based on the solution data, determine the position, shape and size information of the potential defect, and generate defect location data including three-dimensional coordinates and contour features;

[0172] The second processing module 40 is used to extract key defect information of the target metal material based on the defect location data using an edge detection algorithm; wherein the key defect information includes the contour and depth information of the defect;

[0173] The generating module 50 is used to determine the defect type of the target metal material based on the key defect information and generate the detection result information corresponding to the target metal material.

[0174] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0175] It should be understood that determining B based on A does not mean determining B only based on A. B can also be determined based on A and / or other information.

[0176] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0177] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0178] The preferred embodiments of the present invention disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details in detail, nor do they limit the present application to specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can understand and use the present application well. The present application is limited only by the claims and their full scope and equivalents.

Claims

1. A metal material defect identification and analysis method based on multiple detection technology, characterized in that: include: Using a preset microscope device to detect the surface morphology of the target metal material to obtain surface morphology data; wherein the surface morphology data includes: image data of the surface of the target metal material, and size, shape and distribution information of surface defects of the target metal material; The internal structure of the target metal material is detected based on ultrasonic detection technology to generate ultrasonic detection data; wherein the ultrasonic detection data includes: waveform amplitude, frequency characteristics, depth and shape information of the internal defects of the target metal material; Based on the multi-grid method, feature alignment and fusion processing are performed on the surface morphology data and the ultrasonic detection data to generate solution data of ultrasonic wave field propagation; based on the solution data, the position, shape and size information of the potential defect are determined to generate defect location data including three-dimensional coordinates and contour features; Based on the defect location data, using an edge detection algorithm, extract key defect information of the target metal material; wherein the key defect information includes the contour and depth information of the defect; Based on the key defect information, determine the defect type of the target metal material, and generate the detection result information corresponding to the target metal material; The solution data for generating ultrasonic wave field propagation includes: The surface morphology data and the ultrasonic detection data are respectively subjected to layered processing to extract feature information at different scales; Performing alignment processing on the features of the surface topography data and the ultrasonic detection data at each level by using a multi-grid method, wherein the alignment processing includes feature point matching and geometric transformation; After completing the alignment of each level, the alignment results of each level are fused through the grid interpolation method to generate a unified feature distribution model; According to the characteristic distribution model, a mathematical solution model for ultrasonic wave field propagation is established to generate solution data for ultrasonic wave field propagation; The multi-grid method includes: Based on the geometric constraints of the surface morphology of the target metal material, extracting significant feature points and generating a surface morphology feature point set; Generate an internal structure feature point set based on the depth information and waveform amplitude characteristics of the internal structure in the ultrasonic detection data; The surface morphology feature point set and the internal structure feature point set are matched at multiple scales through the morphology constraint registration method.

2. The metal material defect identification and analysis method based on multiple detection technology according to claim 1 is characterized in that: The multi-scale morphology matching includes: The surface morphology feature point set is fitted into a continuous surface using a surface fitting algorithm to generate a constrained surface model; According to the constrained surface model, a deep feature optimization algorithm is used to transform and adjust the internal structure feature point set to achieve global feature alignment; In response to the completion of global feature alignment, the surface morphology features and the ultrasonic detection features are dynamically adjusted based on a local feature consistency optimization algorithm.

3. The metal material defect identification and analysis method based on multiple detection technology according to claim 2 is characterized in that: The data fusion of the alignment results of each level by the grid interpolation method to generate a unified feature distribution model includes: Determining a fusion weight when performing the data fusion based on the distribution characteristics of the surface morphology data and the ultrasonic detection data; Based on the fusion weight, the alignment results of each level are divided into a plurality of interpolation regions; the region types of the interpolation regions include: data-intensive regions and data-sparse regions; Interpolation processing is performed within each of the interpolation areas.

4. The metal material defect identification and analysis method based on multiple detection technology according to claim 3 is characterized in that: Determining the fusion weight includes: Determining the weight of the surface topography data based on the surface defect distribution density of the target metal material; The weight of the ultrasonic detection data is determined based on the depth and signal strength of the internal defects of the target metal material.

5. The metal material defect identification and analysis method based on multiple detection technology according to claim 4 is characterized in that: The interpolation process includes: For the data-intensive area, high-order polynomial interpolation is adopted; For the data sparse area, a distance weighted interpolation method or a low-order interpolation method is adopted.

6. The metal material defect identification and analysis method based on multiple detection technology according to claim 5 is characterized in that: The extracting of key defect information of the target metal material also includes: Based on the defect location data, generating input data for edge detection; wherein the input data for edge detection includes: three-dimensional coordinates, contour features, and depth information of the defect; Using the Canny edge detection algorithm to process the defect area in the input data, and extract the defect edge information of the target metal material; Based on the defect edge information, key defect information of the target metal material is extracted.

7. The metal material defect identification and analysis method based on multiple detection technology according to claim 6 is characterized in that: Before extracting the key defect information of the target metal material based on the defect edge information, the method further includes: optimizing the edge detection result based on the three-dimensional coordinate constraint in the defect location data; wherein the optimizing the edge detection result includes: Identify and remove isolated edge points by calculating the neighborhood consistency score for each edge point; The morphological dilation algorithm is used to connect the discontinuous edge parts in the edge detection results to generate continuous edge information. According to the three-dimensional coordinate constraints in the defect location data, the spatial range of the edge points is defined, and the edge points that deviate from the range are adjusted.

8. A metal material defect recognition and analysis system based on multiple detection technologies, which is used to implement a metal material defect recognition and analysis method based on multiple detection technologies as described in any one of claims 1 to 7, characterized in that: include: A first acquisition module is used to detect the surface morphology of the target metal material using a preset microscope device to obtain surface morphology data; wherein the surface morphology data includes: image data of the surface of the target metal material, and size, shape and distribution information of surface defects of the target metal material; A second acquisition module is used to detect the internal structure of the target metal material based on ultrasonic detection technology to generate ultrasonic detection data; wherein the ultrasonic detection data includes: waveform amplitude, frequency characteristics, depth and shape information of the internal defects of the target metal material; The first processing module is used to perform feature alignment and fusion processing on the surface morphology data and the ultrasonic detection data based on the multi-grid method to generate solution data of ultrasonic wave field propagation; based on the solution data, determine the position, shape and size information of the potential defect, and generate defect location data including three-dimensional coordinates and contour features; A second processing module is used to extract key defect information of the target metal material based on the defect location data using an edge detection algorithm; wherein the key defect information includes the contour and depth information of the defect; A generation module is used to determine the defect type of the target metal material based on the key defect information and generate detection result information corresponding to the target metal material.

Images