Construction method of weld defect grade evaluation system based on feature fusion

Abstract

The present invention provides a weld defect grade evaluation system based on Gabor gray-level co-occurrence matrix feature fusion, relating to the technical field of weld defect grade evaluation. This weld defect grade evaluation system, based on the acceptance criteria for time-of-flight (TOFD) inspection quality, integrates factors affecting weld defect grade, such as geometric dimensions, texture features, and defect category, with geometric feature indicators such as thickness and maximum defect length, texture feature indicators such as energy and entropy, and indicators such as defect location and defect type in weld defect rating, thereby constructing a weld defect grade evaluation index system. This system addresses the current issues of defect grade evaluation using time-of-flight ultrasonic diffraction (TOFD) atlas data, which is primarily manual, highly subjective, inefficient, and lacks unstructured data such as standard defect grade atlases. It achieves the construction of a defect standard database and establishes a quantifiable defect grade evaluation index system.

Description

Technical Field

[0001] The present invention relates to the technical field of weld defect grade evaluation, and in particular to a weld defect grade evaluation system based on Gabor-gray level co-occurrence matrix feature fusion. Background Art

[0002] Welding processes are extensively used in the manufacturing of large-scale equipment in sectors such as energy and power, aerospace, and shipbuilding. The detection and evaluation of weld defects has become a bottleneck restricting the efficient manufacturing of current equipment and a crucial means of ensuring the safe and reliable operation of equipment during service. Time-of-flight ultrasonic diffraction (TOFD) offers advantages such as rich detection information, strong noise immunity, high efficiency, and accurate positioning and quantitative analysis, making it one of the most widely used non-destructive testing methods for welds. For example, the Shaanxi Provincial Special Inspection Institute uses TOFD technology in its special equipment weld inspection and testing, accumulating over 50,000 meters of TOFD weld image data throughout 2020. Currently, defect assessment using data primarily relies on manual methods, referencing standards for defect grading. This involves technicians making judgments based on their own experience and professional background. This process is labor-intensive, subjective, and has inconsistent results, making it difficult to achieve standardized and accurate defect assessment. Therefore, developing a weld defect grading method based on TOFD inspection image data is of great significance for improving defect analysis and determination capabilities and enhancing the safe and serviceable quality of equipment.

[0003] With the gradual application of TOFD inspection technology and the accumulation of inspection data, the use of artificial intelligence to achieve intelligent defect grade assessment, improve defect identification efficiency, and reduce inconsistencies in manual identification has become a key issue. Intelligent and automatic assessment of defect grades in welds inspected for nondestructive testing of carbon steel, low-alloy steel, and especially titanium alloys has long been a technology of great interest to companies. However, very little research has been conducted domestically and internationally. These methods primarily focus on constructing labeled data with defect grades and using existing AI models to classify them. However, as demonstrated by table lookup analysis and image comparison methods in practice, defect grade assessment requires multiple elements, including both structured data (such as rating tables and quantitative descriptions) and unstructured data (such as standard defect grade atlases). Therefore, existing assessment methods based solely on "image classification models" fall far short of industry accuracy requirements. Currently, there are no references domestically or internationally on TOFD defect grade assessment.

[0004] In summary, the current defect grade assessment work mainly uses manual rating methods, which have problems such as low defect grade evaluation efficiency, high labor intensity, high personnel qualification requirements, high subjectivity in the assessment process, and poor consistency of results. The TOFD defect grade assessment method based on artificial intelligence is currently less studied and is not yet mature. The main reason is the lack of visual analysis of national standards and specifications related to weld defect quality. The assessment is based solely on the size of non-destructive testing defects. The assessment process and results are bound to be untrustworthy and unreliable and difficult to apply in enterprises. Therefore, the intelligent defect grade assessment method needs to solve two problems: one is to establish a quantifiable defect grade evaluation index system, and the other is to realize the intelligent matching decision problem of grade mode.

[0005] To address the above problems, we have developed a new weld defect grade evaluation system based on Gabor-gray level co-occurrence matrix feature fusion. Summary of the Invention

[0006] (1) Technical problems solved

[0007] In response to the shortcomings of the existing technology, the present invention provides a weld defect grade evaluation system based on Gabor-gray level co-occurrence matrix feature fusion, which solves the problems in the existing technology that the current defect grade assessment work mainly adopts the manual rating method, which has the disadvantages of low defect grade evaluation efficiency, high labor intensity, high personnel qualification requirements, high subjectivity in the assessment process and poor consistency of results, as well as the fact that the TOFD defect grade assessment method based on artificial intelligence is currently less studied and is not yet mature.

[0008] (2) Technical solution

[0009] To achieve the above objectives, the present invention is implemented through the following technical solutions: a weld defect grade evaluation system based on Gabor-gray level co-occurrence matrix feature fusion, comprising:

[0010] A geometric evaluation index for weld defect grade, which quantitatively reflects the geometric size and shape of the defect based on geometric features, including workpiece thickness, number of defects, defect length, and defect height;

[0011] Texture analysis method, which is based on texture features and is sensitive to changes in image scale and direction;

[0012] The weld defect grade evaluation index, based on the acceptance criteria for TOFD inspection quality, integrates factors such as geometric dimensions, texture characteristics, and defect categories that affect the weld defect grade with geometric characteristic indicators such as detection thickness and maximum defect length, texture characteristic indicators such as energy and entropy, as well as indicators such as defect location and defect type in weld defect rating, thereby constructing a weld defect grade evaluation index system.

[0013] Furthermore, an algorithm for determining geometric features is provided:

[0014] Let the thickness of the workpiece be t, the number of defects in the evaluated defect area in the image be N, and the length of the i-th (0 < i ≤ N) defect in the image be l

[0025] , and its own height be h i ;

[0015] Record the thickness of the detected workpiece as t, the total length of all defects in the selected evaluation area as L, and the total number as N; the largest defect in the selected area is the one with the largest length among all defects. Let it be the j-th (0 < j ≤ N) defect, the length of its minimum circumscribed rectangle be l, and the height of the defect itself be h.

[0016] Furthermore, an algorithm for a texture analysis method based on waveform transformation technology is provided, including:

[0017] The Gabor function can extract relevant features at different scales and in different directions in the frequency domain. The Gabor function is actually a Gaussian function modulated by a complex sine, and its general form is expressed as follows:

[0018]

[0019] In the field of image processing, the two-dimensional Gabor filter is usually used as a linear filter for edge detection, and its basic function expression is as shown in the formula:

[0020] x' = +xcosθ + ysinθ (2)

[0021] y' = -xsinθ + ycosθ (3)

[0022] In the formula: σ x and σ y respectively represent the spatial breadth and bandwidth characteristics of the Gabor basis function, usually describing the size of the filter. Generally, σ x = σ y = σ, where σ is the spatial constant; ω is the spatial frequency of the sine function; θ is the direction for the filter to extract features, and different θ represents features of different angles of the image.

[0023] Furthermore, an algorithm for a texture analysis method based on the gray-level co-occurrence matrix is provided, including:

[0024] Let a given m×n defect image I, whose gray levels can be divided into Q levels, then the calculation of the gray-level co-occurrence matrix P of the image I is as follows:

[0025] p(i,j) = p(I(x,y)) = i, I(x + d cosθ, y + d sinθ) = j) (4)

[0026] Where: p(i,j) is the value of the gray-level co-occurrence matrix at (i,j); (i,j) is a gray-level value pair and i,j∈{0,1,2,·…,Q}; I(x,y) is the gray-level magnitude of the coordinate position (x,y) in image I; d and θ are adjustable parameters, representing the Euclidean distance and the corresponding orientation angle of two coordinate points in the image space, respectively.

[0027] From the above definition, it can be seen that the element value of the gray level co-occurrence matrix P position (i, j) is the distance d and the direction The probability of a pair of pixels with grayscale values ​​​​i and j appearing is a value of Matrix of

[0028] Select d and The gray-level co-occurrence matrix of various distances and angles can be obtained, and 5 commonly used parameters are selected to extract the original feature parameters for defect classification;

[0029] Energy (ASM)

[0030]

[0031] Entropy (ENT)

[0032]

[0033] Correlation (COR)

[0034]

[0035] Where:

[0036] Homogeneity (IDM)

[0037]

[0038] Contrast (CON)

[0039]

[0040] Furthermore, the weld defect grade evaluation index system mainly includes two levels, among which the target layer of the evaluation system represents the weld defect grade evaluation status. The first-level evaluation indicators include three items in total, such as geometric characteristic factors and texture characteristic factors; the second-level evaluation indicators include 14 items in total, such as detection thickness and maximum defect length.

[0041] (3) Beneficial effects

[0042] The present invention provides a weld defect grade evaluation system based on Gabor-gray level co-occurrence matrix feature fusion, which has the following beneficial effects:

[0043] This weld defect grade evaluation system based on Gabor-grayscale co-occurrence matrix feature fusion, based on the acceptance criteria for TOFD inspection quality, integrates factors affecting weld defect grade, such as geometric dimensions, texture features, and defect categories, with geometric feature indicators such as detection thickness and maximum defect length, texture feature indicators such as energy and entropy, as well as indicators such as defect location and defect type in weld defect rating, to construct a weld defect grade evaluation index system. This solves the current problems of defect grade assessment using ultrasonic time-of-flight diffraction (TOFD) atlas data, which is mainly manual assessment, highly subjective, inefficient, and lacks unstructured data such as standard defect grade atlases. It realizes the construction of a defect standard database and establishes a quantifiable defect grade evaluation index system. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a typical surface opening defect spectrum of the weld of the present invention;

[0045] Figure 2 This is a typical planar defect spectrum of the weld of the present invention;

[0046] Figure 3 The original image is the effect of the Gabor filter of the present invention on extracting edge features of an unfused image;

[0047] Figure 4 The Gabor filter of the present invention is used to extract the edge features of the unfused image. Figure 1 ;

[0048] Figure 5 The Gabor filter of the present invention is used to extract the edge features of the unfused image. Figure 2 ;

[0049] Figure 6 The Gabor filter of the present invention is used to extract the edge features of the unfused image. Figure 3 ;

[0050] Figure 7 The Gabor filter of the present invention is used to extract the edge features of the unfused image. Figure 4 ;

[0051] Figure 8 A pixel pair image of the gray-level co-occurrence matrix diagram of the present invention;

[0052] Figure 9 Schematic diagram of the TOFD weld defect grade evaluation index system of the present invention;

[0053] Figure 10 Schematic diagram of the evaluation indexes for different weld defect types according to the present invention. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0055] Example

[0056] In the existing technology, ultrasonic time-of-flight diffraction (TOFD) has the advantages of rich detection information, strong noise resistance, high efficiency, and accurate positioning and quantitative detection. It has become one of the most widely used non-destructive testing methods for welds. Specifically, TOFD testing is a method that can obtain diffraction energy from the "end angles" and "end points" of the internal defects of the test piece to be inspected to detect defects.

[0057] According to the NB / T47013.10-2015 "Non-destructive testing of pressure equipment Part 10: Time-of-flight diffraction ultrasonic testing" standard and the BS7706 standard, TOFD defect classification is based on the echo characteristics and phase state of the emitter to characterize the texture characteristics, signal-to-noise ratio, and morphological characteristics of the defects.

[0058] The defect level of upper surface opening, lower surface opening, cracks in buried defects and lack of fusion is level III, which has nothing to do with the detection thickness. Other types of defects have a strong correlation with the detection thickness. The defects of upper surface opening, lower surface opening, cracks in buried defects and lack of fusion are applicable to all detection thicknesses. The detection thickness of other defect types is based on the engineering thickness detected in the standard atlas, such as Figure 1 、 Figure 2 The standard maps of typical surface opening defects and planar defects are shown.

[0059] Different types of defects have distinct geometric and textural characteristics. Correctly understanding these characteristics is fundamental to defect grading by engineers. Therefore, this paper, in conjunction with relevant grading standards, comprehensively considers factors such as geometric size, quantity, distribution, and the texture characteristics of the defects themselves. It quantifies weld defects using geometric and textural characteristics, establishes a mapping between actual defects and standard requirements, and constructs a defect grading index system.

[0060] like Figure 3 - Figure 10 As shown, the embodiment of the present invention provides a weld defect grade evaluation system based on Gabor-gray level co-occurrence matrix feature fusion, including:

[0061] Geometric evaluation index for weld defect grades, where the geometric evaluation index for weld defect grades quantitatively reflects the geometric dimensions and shape characteristics of defects based on geometric features, and the geometric features include workpiece thickness, number of defects, defect length, and defect height;

[0062] Texture analysis method, which is sensitive to changes in image scale and direction based on texture features;

[0063] Weld defect grade evaluation index. According to the acceptance criteria for TOFD detection quality, factors such as geometric dimensions, texture features, and defect categories that affect weld defect grades are integrated with geometric feature indexes such as detection thickness and maximum defect length, texture feature indexes such as energy and entropy, and indexes such as defect location and defect type in weld defect rating, thereby constructing a weld defect grade evaluation index system.

[0064] Among them:

[0065] 1) Geometric evaluation index for weld defect grades

[0066] Geometric features are used to quantitatively reflect the geometric dimensions and shape characteristics of defects, which are of crucial significance in the evaluation of titanium alloy weld defects, mainly including workpiece thickness, number of defects, defect length, and defect height.

[0067] Provide an algorithm for determining geometric features: Let the workpiece thickness be t, the number of defects in the evaluated defect area in the image be N, and the length of the i-th (0 < i ≤ N) defect in the image be l i , and its own height be h i .

[0068] Record the detected workpiece thickness as t, the total length of all defects in the selected evaluation area as L, and the total number as N; the largest defect in the selected area is the one with the largest length among all defects. Let it be the j-th (0 < j ≤ N) defect, the length of its minimum circumscribed rectangle be l, and the height of the defect itself be h;

[0069] Then the geometric features are defined as shown in Table 1:

[0070] Table 1 Definition of geometric features

[0071]

[0072] 2) Texture analysis method based on waveform transformation technology

[0073] Texture is an important image feature, and texture features are sensitive to changes in image scale and direction. Wavelet transform is a new theory for analyzing texture features that has developed rapidly in recent years and is a method based on signal filtering.

[0074] An algorithm for a texture analysis method based on waveform transformation technology is provided, comprising:

[0075] The Gabor function can extract relevant features at different scales and directions in the frequency domain. The Gabor function is actually a Gaussian function modulated by a complex sine wave. Its general form is as follows:

[0076]

[0077] Gabor filters are widely used in the fields of visual information processing and image understanding. In the field of image processing, two-dimensional Gabor filters are usually used as linear filters for edge detection. Its basic function expression is shown as follows:

[0078] x'=+xcosθ+ysinθ (2)

[0079] y'=-xsinθ+ycosθ (3)

[0080] Where: σ x and σ y Respectively characterize the spatial breadth and bandwidth characteristics of the Gabor basis function, usually describing the size of the filter. In general, σ x =σ y =σ, where σ is the spatial constant; ω is the spatial frequency of the sine function; θ is the direction in which the filter extracts features, and different θ represents features at different angles of the image.

[0081] from Figure 3 It can be seen that the original Figure 3 The edges of the unfused defects are mainly horizontal or approximately horizontal. Since the Gabor filter can extract edge features in the original image that are perpendicular to the Gabor feature extraction direction, when the feature extraction direction is 0°, 45°, and 135°, the output image fails to retain the horizontal edge features of the unfused defects. Figure 4 、 5 When the feature extraction direction is 90°, the output image retains the horizontal edge features well, as shown in Figure 6 As shown in the figure, the Gabor filter has the ability to extract edge features in the weld defect image that are perpendicular to the Gabor kernel feature extraction direction, and can better preserve edge information in a specific direction in the image. However, the TOFD morphological features of weld defects are uncertain and multi-directional. In order to better characterize the texture features of the defects, the wavelet coefficient modulus mean and standard deviation normalized for the output images with extraction directions of 0°, 45°, 90°, and 135° are used as feature vectors.

[0082] 3) Texture analysis method based on gray-level co-occurrence matrix

[0083] Texture is an important image feature. It is a pattern formed by the grayscale variations of image pixels. Image texture characteristics are closely related to the grayscale patterns of pixels and their surroundings. For grayscale images, texture reflects the characteristics of the grayscale distribution in the global image. Based on this characteristic, Haralick et al. proposed the Gray Level Co-occurrence Matrix (GLCM) to describe the texture variations of images.

[0084] like Figure 8 The gray-level co-occurrence matrix studies the spatial correlation characteristics of image grayscale by counting the frequency of occurrence of grayscale values ​​of pixel pairs with a certain specific position relationship in space. It is an estimate of the joint probability density function of grayscale pairs in the image and an expression of comprehensive information such as the direction, change amplitude and grayscale distribution of the local neighborhood of the image.

[0085] An algorithm for a texture analysis method based on a gray-level co-occurrence matrix is ​​provided, comprising:

[0086] Assume a given m×n defect image I, whose grayscale can be divided into Q levels, then the grayscale co-occurrence matrix P of image I is calculated as follows:

[0087] p(i,j)=p(I(x,y))=i,I(x+d cosθ,y+d sinθ)=j) (4)

[0088] Where: p(i,j) is the value of the gray-level co-occurrence matrix at (i,j); (i,j) is a gray-level value pair with i,j∈0,1,2,…,Q}; I(x,y) is the gray-level magnitude of the coordinate position (x,y) in image I; d and θ are adjustable parameters, representing the Euclidean distance and the corresponding orientation angle of two coordinate points in the image space, respectively.

[0089] From the above definition, it can be seen that the element value of the gray level co-occurrence matrix P position (i, j) is the distance d and the direction The probability of a pair of pixels with grayscale values ​​​​i and j appearing is a value of Matrix of

[0090] Select d and The gray-level co-occurrence matrix of various distances and angles can be obtained, and 5 commonly used parameters are selected to extract the original feature parameters for defect classification;

[0091] Energy (ASM)

[0092]

[0093] Entropy (ENT)

[0094]

[0095] Correlation (COR)

[0096]

[0097] Where:

[0098] Homogeneity (IDM)

[0099]

[0100] Contrast (CON)

[0101]

[0102] The above introduction to the gray-level co-occurrence matrix and its statistics demonstrates that the statistics calculated using the gray-level co-occurrence matrix in TOFD images can effectively describe texture features in terms of spatial distribution, something that cannot be achieved using the Gabor kernel in wavelet transforms. However, the gray-level co-occurrence matrix lacks detailed texture information in its description, which can be compensated for by the Gabor kernel. Therefore, combining the advantages of the gray-level co-occurrence matrix in terms of spatial distribution with the advantages of the Gabor kernel in terms of local structural detail creates a joint texture feature that can better describe the texture characteristics of TOFD images.

[0103] 4) Weld defect grade evaluation index

[0104] Based on the analysis of weld grade evaluation indicators, this paper proposes the following Figure 9 The weld defect grade evaluation factor system shown in the figure is based on the acceptance criteria for TOFD inspection quality. The core concept is to integrate factors that affect weld defect grade, such as geometric dimensions, texture characteristics, and defect categories, with geometric characteristic indicators such as thickness and maximum defect length, texture characteristic indicators such as energy and entropy, and indicators such as defect location and defect type in weld defect rating, thereby constructing a weld defect grade evaluation index system.

[0105] Figure 9 The weld defect grade evaluation index system shown mainly includes two levels, among which the target layer of the evaluation system represents the weld defect grade evaluation status. The first-level evaluation indicators include three items in total, such as geometric characteristic factors and texture characteristic factors; the second-level evaluation indicators include 14 items in total, such as detection thickness and maximum defect length.

[0106] According to the above analysis, combined with the experience in actual enterprise production, according to the above definition, the quantitative index of multi-source information fusion weld defect grade assessment is constructed as follows: Figure 10 .

[0107] In this paper, in order to address the problems of current defect grade assessment using time-of-flight ultrasonic diffraction (TOFD) atlas data, which is mainly based on manual assessment, is highly subjective, inefficient, and lacks unstructured data such as standard defect grade atlases, a weld defect grade evaluation system based on Gabor-grayscale co-occurrence matrix feature fusion is proposed. By analyzing the characteristics of TOFD defect detection data and based on detection standards and expert domain knowledge, an evaluation system with multi-source information fusion is established, thus realizing the construction of a defect standard database.

[0108] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a weld defect grade evaluation system based on feature fusion, characterized in that: Including: Constructing geometric evaluation indexes for weld defect grades, where the geometric evaluation indexes for weld defect grades quantitatively reflect the geometric dimensions and shape characteristics of defects based on geometric features, and the geometric features include workpiece thickness, number of defects, defect length, and defect height; Constructing texture features through texture analysis methods, and the construction of texture features includes: constructing first texture features using a texture analysis method based on waveform transformation technology; constructing second texture features using a texture analysis method based on gray-level co-occurrence matrix; obtaining the texture features through the first texture features and the second texture features, and the texture features are joint texture features; the texture features are sensitive to changes in image scale and direction; the texture analysis method based on waveform transformation technology includes: Extracting relevant features at different scales and different directions in the frequency domain through the Gabor function, and the Gabor function is a Gaussian function modulated by a complex sine, and its form is expressed as follows: In the field of image processing, using a two-dimensional Gabor filter as a linear filter for edge detection, and its basic function expression is as shown in the formula: x' = +xcosθ + ysinθ (2) y' = -xsinθ + ycosθ (3) Where: σ x and σ y Respectively characterize the spatial breadth and bandwidth characteristics of the Gabor basis function and describe the size of the filter, σ x =σ y =σ, where σ is the spatial constant; ω is the spatial frequency of the sine function; θ is the direction in which the filter extracts features, and different θ represents features at different angles of the image; The texture analysis method based on gray-level co-occurrence matrix includes: Given a defect image I of m×n, whose gray levels are divided into Q levels, the calculation of the gray-level co-occurrence matrix P of the image I is as follows: p(i,j) = p(I(x,y)) = i, I(x + d cosθ, y + d sinθ) = j) (4) In the formula: p(i,j) is the value of the gray-level co-occurrence matrix at (i,j); (i,j) is a pair of gray values and i,j ∈ {0,1,2,…,Q}; I(x,y) is the magnitude of the gray level at the position of coordinates (x,y) in the image I; d and θ are adjustable parameters, representing the Euclidean distance between two coordinate points in the image space and the corresponding azimuth angle respectively; The element value of the gray level co-occurrence matrix P position (i, j) is the distance d and the direction The probability of a pair of pixels with grayscale values ​​​​i and j appearing is a value of Matrix of Select d and The gray-level co-occurrence matrix of various distances and angles can be obtained, and 5 commonly used parameters are selected to extract the original feature parameters for defect classification; Energy ASM Entropy ENT Correlation COR Where: Homogeneity IDM Contrast CON 2. The method for constructing a weld defect grade evaluation system based on feature fusion according to claim 1, characterized in that: According to the acceptance criteria for TOFD detection quality, integrating the geometric feature indexes, texture feature indexes, defect position and defect type indexes that affect the weld defect grade, and then constructing a weld defect grade evaluation index system. Let the thickness of the workpiece be t, the number of defects in the evaluated defect area in the image be N, and the length of the i-th (0 < i ≤ N) defect in the image be l i , and its own height be h i ; Determining the geometric features includes:

3. The method for constructing a weld defect grade evaluation system based on feature fusion according to claim 1, characterized in that: Denote the thickness of the detected workpiece as t, the total length of all defects in the selected evaluation area as L, and the total number as N; the largest defect in the selected area is the one with the largest length among all defects. Let it be the jth (0 < j ≤ N) defect, the length of its minimum circumscribed rectangle is l, and the height of the defect itself is h. The weld defect grade evaluation index system includes two levels, where the target layer of the evaluation system represents the weld defect grade evaluation status, and the first level includes geometric feature factors and texture feature factors; The second level includes detection thickness and maximum defect length factors.

Images