Brightness combination optimization method for predicting alloy segregation element content distribution

Abstract

The invention discloses a brightness combination optimization method for predicting alloy segregation element content distribution. The method comprises the following steps: displaying an alloy structure based on an acid etching reaction to obtain an alloy structure morphology; changing the brightness of a light source, and obtaining tissue morphology images under different brightness conditions by using a photographing device; analyzing image quality characteristics and gray level distribution of different regions of the tissue, and determining the optimal image brightness beneficial to gray level analysis of regions with different segregation element contents in the tissue image; the segregation element content is divided into a high-content area, a middle-content area and a low-content area from high to low, the segregation element content is recombined with the gray values of the tissue images under different brightness, and a correlation model is constructed; and calculating segregation element content distribution of an unmeasured area or a similar new sample based on the grey-scale maps of the tissue photos under different brightness and a correlation model. By means of the method, the correlation between the alloy segregation element content and the structure image gray level can be improved, and therefore the alloy large-area multi-scale segregation element content can be obtained efficiently.

Description

Technical Field

[0001] The invention relates to the field of alloy segregation element content distribution detection and quality stability control, and in particular to a brightness combination optimization method for predicting alloy segregation element content distribution. Background Art

[0002] Taking carbon in steel as an example, due to its light weight and small solute distribution coefficient, it is easy to form uneven carbon content in different regions during the solidification of molten steel, which in turn causes carbon segregation. Carbon segregation is a common internal defect of carbon steel continuous casting billets, which generally directly affects the quality of continuous casting billets and final products. In production process optimization and scientific research, it is often necessary to use various technical means to detect and analyze the distribution of carbon content in carbon steel continuous casting billets, evaluate the uniformity of element distribution and corresponding organization, and provide reference for process improvement. The existing methods for measuring the carbon content of carbon steel continuous casting billets mainly include chemical analysis, electron probe and other methods. However, they have the problems of needing to sample cuttings first, being highly destructive to the sample, time-consuming and labor-intensive operation, high detection cost, small measurement range, and needing to excite the sample surface during the measurement process and the excitation result is a statistical average value in a certain area (the analysis scale is generally millimeter level), so that the measurement results obtained after the measurement cannot correspond well to the low-magnification organization (morphology such as solidification organization and segregation) and the specific detection position. At present, high-end steel materials have increasingly higher requirements for the uniformity of quality in large areas of steel. There is an urgent need for a method that can carry out a precise and comprehensive detection of the segregation of carbon elements in carbon steel continuous casting billets. Therefore, it is necessary to be able to carry out a more efficient detection of the distribution of carbon content in carbon steel casting billets, that is, to be able to meet the analysis requirements of large areas at the same time, and to ensure the detection accuracy, and to be simple to operate, low cost, low destructiveness, and high efficiency. The existing detection methods can no longer meet the requirements.

[0003] The grayscale analysis method of metal macrostructure (abbreviated as grayscale analysis method) was proposed by Hou Zibing et al. in "Method for obtaining C content distribution using grayscale image of carbon steel ingot" (published in Journal of Iron and Steel Research, Vol. 31, No. 7, pp. 620 to 627, 2019) and "Method for obtaining carbon content distribution using grayscale image of carbon steel ingot" (patent number CN201610405221.6, authorization announcement number CN105866124B, 2018). The method is to establish a carbon content prediction model based on the special relationship between the grayscale of the macrostructure image of the continuous casting ingot and the carbon content at the corresponding position, and then realize the carbon content characterization based on the grayscale of the macrostructure image of the continuous casting ingot. Since the process of obtaining the macrostructure image and its grayscale information is simple to operate, low-cost, and the pixel scale can be at the micrometer or millimeter level, the grayscale analysis method of metal macrostructure has the potential to accurately and efficiently detect the degree of carbon segregation in a large area of ​​the entire continuous casting ingot sample. However, due to the large differences in the microstructure of different carbon content areas in the continuous casting billet, the sensitivity to different brightness and different position image acquisition is different. The specific image acquisition conditions should be considered when using the metal low-magnification structure grayscale analysis method. This method analyzes the image quality characteristics and grayscale distribution of the overall and local areas of the alloy structure images obtained under different brightness conditions, and proposes a brightness combination optimization method, that is, the grayscale combination of the structure images under different brightness in different segregation element areas is combined, and then a function model is established with the segregation element content at the corresponding position. This can further improve the accuracy of the segregation element content results obtained based on grayscale analysis, and its application range is wider. Summary of the invention

[0004] The present invention aims to provide a brightness combination optimization method (Brightness combination optimization, BCO for short) for predicting the distribution of alloy segregation element content. This method is used to improve the correlation between the alloy element content at the corresponding position and the grayscale of the alloy structure image, thereby improving the accuracy of the grayscale analysis results, laying a solid foundation for achieving precise and efficient detection of the segregation degree of a large area of ​​the entire alloy sample.

[0005] The technical solution of the present invention is a brightness combination optimization method for predicting the distribution of alloy segregation element content, characterized in that it includes the following steps:

[0006] (1) Display of sample cross-section microstructure: According to the process requirements, the cross-section of the alloy microstructure sample is acid-etched to display the morphology of the cross-section of the alloy microstructure sample;

[0007] (2) Obtaining sample cross-section microstructure images: Using photographic equipment to take photos of the sample cross-section under different light source brightness conditions, obtain low-magnification microstructure morphology images of the cross-section of the continuous casting billet sample after acid etching;

[0008] (3) Analysis of image quality characteristics and grayscale distribution of tissues at different brightness: Select the complete area including the starting and ending positions of segregation for electron probe or chemical analysis to obtain the one-dimensional or two-dimensional actual content distribution of segregation elements, analyze the image quality characteristics and grayscale distribution of low-magnification tissues through the image information entropy of the cross-sectional tissue, and / or combine the fitting results of the grayscale distribution and the actual content distribution of the segregation elements to determine the optimal brightness M for predicting the content distribution of segregation elements based on the overall tissue image; then use the same method to analyze the area with high content of segregation elements to obtain the optimal brightness L of the corresponding area; then use the same method to analyze the area with low content of segregation elements to obtain the optimal brightness H of the corresponding area;

[0009] (4) Grayscale combination and model construction of images with different brightness: sort the segregation element content from high to low, match and sort the grayscale values ​​of the three different brightness levels of L, M, and H with the segregation element content, and then divide the segregation element content from high to low into three intervals: high content area, medium content area, and low content area, and obtain the dividing element content C1 of the high content area and the medium content area and the dividing element content C2 of the medium content area and the low content area; then for the high content area, take the grayscale value under L brightness; for the medium content area, take the grayscale value under M brightness; for the low content area, take the grayscale value under H brightness, and thus obtain the grayscale critical value G1 that distinguishes the high content area and the medium content area and the grayscale critical value G2 that distinguishes the medium content area and the low content area; based on the above values, obtain the grayscale combination distribution that matches the segregation element content distribution after combination; based on the grayscale combination distribution and the segregation element content distribution, construct a correlation model between grayscale and element content by fitting equations.

[0010] (5) Model application: First, for other unmeasured areas or new samples of the same type, the image grayscale values ​​corresponding to the three different brightness levels of L, M, and H are obtained and sorted from small to large. The sorted grayscale values ​​are divided into three intervals using G1 and G2 as the dividing values. Then, the grayscale value less than G1 is taken at L brightness, the grayscale value between G1 and G2 is taken at M brightness, and the grayscale value greater than G2 is taken at H brightness, thereby obtaining the combined grayscale distribution. Then, the correlation model obtained in the previous step is used to calculate and predict the distribution of segregation element content in other unmeasured areas or new samples of the same type.

[0011] Furthermore, the etching liquid is a hydrochloric acid alcohol aqueous solution with a volume ratio of 1.0: (0.1-0.5): (0.5-2.0), or a hydrochloric acid sulfuric acid alcohol aqueous solution with a volume ratio of 3.0: (0.5-1.0): (0.1-0.5): (0-4.0); or a hydrochloric acid nitric acid alcohol solution with a volume ratio of (0-5.0): (0.5-1.0): (5.0-15.0), and the heating temperature of the solution in the etching tank is 20-70°C. The process conditions of low-frequency vibration etching are: the vibration frequency of the ultrasonic etching device is set to 20-50KHz, the power density is 0.5-1.0W / cm 2 , the acid etching time is 5 to 20 minutes; the conventional tissue acid etching method can refer to the corresponding national standards. For example, for steel, refer to the national standard GB / T226-2015 "Macrostructure and Defect Acid Etching Test Method for Steel".

[0012] Furthermore, except for the brightness of the light source, other parameters of the image acquisition process (the distance between the sample and the light source, the image acquisition range, etc.) are the same, and different color or grayscale of the image can reflect the illumination intensity of different colors of light in the light source.

[0013] Furthermore, the image information entropy of the cross-section structure under different brightness conditions and the image grayscale distribution of the segregation element content area at different positions are obtained, and compared with the actual ingot segregation element distribution. The image information entropy is calculated by the formula:

[0014]

[0015] Where l is the color level of the pixel in the tissue image; L is the maximum color level, L = 255; P(l) is the proportion of pixels with color level l in the tissue morphology image, and E is the image information entropy.

[0016] The grayscale distribution of the image is compared with the actual segregation element content. The determination coefficient R of the result is calculated by fitting the equation 2 To judge, R 2 The larger the value, the better the fitting effect. 2 The smaller it is, the worse the fitting effect is.

[0017]

[0018] Where: u i is the measured value; is the model prediction value; is the average of the measured values.

[0019] Furthermore, the fitting equation and correlation model are mainly fitted and constructed using the following equation coefficients.

[0020]

[0021] Where: C i is the content of segregation elements at a certain position of the alloy calculated by the model; is the mean value of alloy segregation element content; n is the correction coefficient of the mean value of ingot segregation element content; k0 is the solute equilibrium distribution coefficient of the segregation element; a is the correction coefficient of the solute equilibrium distribution coefficient of the segregation element; g i The gray value of the tissue image at a certain position point of the billet; is the grayscale mean in the grayscale image of the ingot structure; b is the correction coefficient of the ratio parameter of the grayscale value at a certain position in the grayscale image of the ingot structure to the grayscale mean.

[0022] Through multiple studies, the inventor team of this patent found that in addition to the high element content, the segregation area also has a more complex microscopic metallographic structure. When the brightness is high, due to the diffuse reflection of the complex microstructure, the grayscale distribution fluctuation range of the segregation area is larger. The principle that the carbon content is negatively correlated with the grayscale of the low-magnification tissue morphology image is that the higher the content of the segregation element, the faster the corrosion rate during the acid etching process, the darker the morphology image after corrosion, and the lower the grayscale. Therefore, when the brightness is high, the diffuse reflection of the complex micromorphology of the segregation area will interfere with the correspondence between the segregation element content and the grayscale of the tissue morphology at the corresponding position. Therefore, combined with the actual situation of the comprehensive alloy sample and the negative correlation between the segregation element content and the tissue morphology image at the corresponding position, the carbon content of the segregation area can better correspond to the image morphology when the brightness is lower than the brightness when the overall correspondence between the segregation element content and the grayscale of the low-magnification tissue morphology image is the best. On the contrary, in the low-content area, the element content of the segregation area can better correspond to the image morphology when the brightness is higher than the brightness when the overall correspondence between the segregation element content and the grayscale of the tissue morphology image is the best.

[0023] In summary, the brightness combination optimization method for predicting the distribution of alloy segregation element content of the present invention has the following advantages: it can improve the correlation between the alloy element content at the corresponding position of the continuous casting billet and the grayscale of the microstructure morphology image, reduce the more discrete points in the scatter diagram of the segregation element content and the grayscale fitting of the microstructure image, thereby reducing the error of the fitting result, establishing a more accurate prediction model for the segregation element content, and improving the accuracy of the grayscale analysis results; at the same time, since the acid etching test is relatively simple, it is easy to obtain the microstructure morphology of a large area and the image pixels are easy to perform multi-scale calculations, the brightness microstructure optimization method is easy to achieve multi-scale characterization of the overall large-area segregation element content of the alloy sample, thereby providing a basis for fine and efficient detection of alloy uniformity and stable quality control. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 Low-magnification tissue images (1.2mm×1.2mm) of a typical location of high-carbon steel (70#) obtained under different brightness conditions;

[0025] Figure 2 The variation curve of the carbon content of high carbon steel (70#) and the fitting determination coefficient and information entropy of the image grayscale function with the increase of brightness;

[0026] Figure 3 Scatter plot and fitting curve of segregated carbon content and gray value of high carbon steel (70#) after brightness combination optimization;

[0027] Figure 4 The distribution diagram of segregated carbon element content in high carbon steel (70#) continuous casting billet with a cross-sectional size of 150mm×150mm calculated based on the brightness combination optimization method. DETAILED DESCRIPTION

[0028] The basic invention is described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0029] The brightness combination optimization method for predicting the content distribution of alloy segregation elements of the present invention is characterized by comprising the following steps:

[0030] (1) Sample cross-section microstructure morphology: According to the process requirements, the cross-section of the alloy microstructure sample is acid-etched to display the morphology of the cross-section of the alloy microstructure sample and obtain a morphology that meets the process requirements;

[0031] The acid etching treatment is based on the combined effect of ultrasound and acid etching reaction, or the acid etching treatment method of the prior art; before the acid etching treatment, if necessary, the sample section is polished, ground, etc. to achieve the required surface finish;

[0032] The ultrasonic and acid etching reaction combined treatment is performed by setting the vibration frequency of the ultrasonic acid etching device to 20-50KHz and the power density to 0.5-1.0W / cm 2 , the acid etching time is 5 to 20 minutes; the acid etching treatment of the prior art can refer to the corresponding national standards. For example, for steel, reference can be made to the national standard GB / T 226-2015 "Test Method for Macrostructure and Defect Acid Etching of Steel".

[0033] The acid etching solution of the acid etching treatment is a hydrochloric acid alcohol aqueous solution with a volume ratio of 1.0: (0.1-0.5): (0.5-2.0), or a hydrochloric acid sulfuric acid alcohol aqueous solution with a volume ratio of 3.0: (0.5-1.0): (0.1-0.5): (0-4.0); or a hydrochloric acid nitric acid alcohol solution with a volume ratio of (0-5.0): (0.5-1.0): (5.0-15.0), and the solution in the acid etching tank is heated to a temperature of 20-70°C.

[0034] (2) Obtaining sample cross-section microstructure images: Using photographic equipment to take photos of the sample cross-section under different light source brightness conditions, obtain low-magnification microstructure morphology images of the cross-section of the continuous casting billet sample after acid etching;

[0035] Except for the brightness of the light source, other parameters including the distance between the sample and the light source, the image acquisition range, etc. are the same during the image acquisition process. The different color or grayscale of the image reflects the illumination intensity of different colors of light in the light source.

[0036] (3) Analysis of image quality characteristics and grayscale distribution of tissues at different brightness: Select the complete area including the starting and ending positions of segregation for electron probe or chemical analysis to obtain the one-dimensional or two-dimensional actual content distribution of segregation elements, analyze the image quality characteristics and grayscale distribution of low-magnification tissues through the image information entropy of the cross-sectional tissue, and / or combine the fitting results of the grayscale distribution and the actual content distribution of the segregation elements to determine the optimal brightness M for predicting the content distribution of segregation elements based on the overall tissue image; then use the same method to analyze the area with high content of segregation elements to obtain the optimal brightness L of the corresponding area; then use the same method to analyze the area with low content of segregation elements to obtain the optimal brightness H of the corresponding area;

[0037] (4) Grayscale combination and model construction of images with different brightness: sort the segregation element content from high to low, match and sort the grayscale values ​​of the three different brightness levels of L, M, and H with the segregation element content, and then divide the segregation element content from high to low into three intervals: high content area, medium content area, and low content area, and obtain the dividing element content C1 of the high content area and the medium content area and the dividing element content C2 of the medium content area and the low content area; then for the high content area, take the grayscale value under L brightness; for the medium content area, take the grayscale value under M brightness; for the low content area, take the grayscale value under H brightness, and thus obtain the grayscale critical value G1 that distinguishes the high content area and the medium content area and the grayscale critical value G2 that distinguishes the medium content area and the low content area; based on the above values, obtain the grayscale combination distribution that matches the segregation element content distribution after combination; based on the grayscale combination distribution and the segregation element content distribution, construct a correlation model between grayscale and element content by fitting equations.

[0038] (5) Model application: First, for other unmeasured areas or new samples of the same type, the image grayscale values ​​corresponding to the three different brightness levels of L, M, and H are obtained and sorted from small to large. The sorted grayscale values ​​are divided into three intervals using G1 and G2 as the dividing values. Then, the grayscale value less than G1 is taken at L brightness, the grayscale value between G1 and G2 is taken at M brightness, and the grayscale value greater than G2 is taken at H brightness, thereby obtaining the combined grayscale distribution. Then, the correlation model obtained in the previous step is used to calculate and predict the distribution of segregation element content in other unmeasured areas or new samples of the same type.

[0039] The image information entropy of the cross-sectional structure under different brightness conditions and the image grayscale distribution of the segregation element content areas at different positions are obtained, and compared with the actual ingot segregation element distribution; the image information entropy is calculated by formula (1):

[0040]

[0041] Where l is the color level of the pixel in the tissue image; L is the maximum color level, L = 255; P(l) is the proportion of pixels with color level l in the tissue morphology image, and E is the image information entropy.

[0042] The grayscale distribution of the image is compared with the actual segregation element content. The determination coefficient R of the result is calculated by fitting the equation 2 To judge, R 2 The larger the value, the better the fitting effect. 2 The smaller it is, the worse the fitting effect is; 2 Calculated by formula (2):

[0043]

[0044] Where: y i is the measured value; is the model prediction value; is the average of the measured values.

[0045] The fitting equation and correlation model are mainly fitted and constructed using the following equation coefficients:

[0046]

[0047] Where: C i is the content of segregation elements at a certain position of the alloy calculated by the model; is the mean value of alloy segregation element content; n is the correction coefficient of the mean value of ingot segregation element content; k0 is the solute equilibrium distribution coefficient of the segregation element; a is the correction coefficient of the solute equilibrium distribution coefficient of the segregation element; g i The gray value of the tissue image at a certain position point of the billet; is the grayscale mean in the grayscale image of the ingot structure; b is the correction coefficient of the ratio parameter of the grayscale value at a certain position in the grayscale image of the ingot structure to the grayscale mean

[0048] Example: Brightness combination optimization method for segregation carbon content distribution of high carbon steel continuous casting slab

[0049] (1) Microstructure of the sample cross section: Taking high carbon steel (70#) with a carbon content of 0.7% as an example, the macrostructure of the 70 steel continuous casting billet is displayed based on the comprehensive effect of ultrasonic and acid etching reaction. The acid etching solution is a hydrochloric acid alcohol aqueous solution with a volume ratio of 1.0:0.5:1. The solution in the acid etching tank is heated to 70°C. During the acid etching process, the acid etching container is placed in an ultrasonic cleaning machine. The process conditions of low-frequency vibration acid etching are: the vibration frequency of the ultrasonic acid etching device is set to 40KHz and the power density is 0.7W / cm 2 , the acid etching time is 10 minutes, and after washing with water, the sample is taken out and dried to obtain the sample;

[0050] (2) Obtaining the image of the sample cross section structure: Using a photographing device, photographs were taken at camera light source brightness of 10%, 15%, 20%, 25%, 30%, 35%, 40%, and 45% to obtain low-magnification structural morphology images of the cross section of the continuous casting billet sample after acid etching (e.g. Figure 1 shown);

[0051] (3) Analysis of image quality characteristics and grayscale distribution of tissues at different brightness: The image brightness that is most conducive to grayscale analysis of carbon content areas in low-magnification tissue images is determined by analyzing the overall image quality characteristics and grayscale distribution of low-magnification tissue images. Figure 2 As shown, due to the information entropy and its determination coefficient R fitting with the actual segregation element content distribution when the brightness is 25% 2 The maximum value is obtained, so the optimal brightness M for predicting the distribution of segregation element content based on the overall tissue image is determined to be 25%; the same method is used to analyze the area with high segregation element content (segregation area), and the optimal brightness L of the corresponding area is 15%; then the same method is used to analyze the area with low segregation element content (dendrite trunk area), and the optimal brightness H of the corresponding area is 35%;

[0052] (4) Grayscale combination and model construction of images with different brightness: sort the segregation element content from high to low, match and sort the grayscale values ​​of the three different brightness levels of 15%, 25%, and 35% with the segregation element content, and then divide the segregation element content into three intervals from high to low: high content area, medium content area, and low content area; then for the high content area, take the grayscale value at 15% brightness; for the medium content area, take the grayscale value at 25% brightness; for the low content area, take the grayscale value at 35% brightness; and thus obtain the corresponding grayscale critical value 85 for distinguishing the high content area and the medium content area and the corresponding grayscale critical value 215 for distinguishing the medium content area and the low content area; based on the above values, obtain the grayscale combination distribution that matches the segregation element content distribution after combination; based on the grayscale combination distribution and the segregation element content distribution, construct a correlation model between grayscale and element content by fitting equations. The fitting coefficient n in the correlation model formula (3) is 0.86, a is 5.7, and b is -3.16; the corresponding figure with the grayscale combination distribution is as follows Figure 3 As shown, the determination coefficient R2 is increased from 0.332 before the brightness combination to 0.622 now. Of course, for alloy structures with different components, the values ​​of L, M, H and the fitting coefficients are different and should be selected according to actual conditions.

[0053] (5) Model application: For a new large-size sample of the same steel grade (the cross-sectional size of the 70 steel continuous casting billet is 150 mm × 150 mm), the image grayscale values ​​corresponding to three different brightness levels of 15%, 25%, and 35% are obtained and sorted from small to large. The sorted grayscale values ​​are divided into three intervals using the grayscale critical values ​​of 85 and 215 as the dividing values. Then, the grayscale values ​​less than 85 are taken at 15% brightness, the grayscale values ​​between 85 and 215 are taken at 25% brightness, and the grayscale values ​​greater than 215 are taken at 35% brightness, thereby obtaining the combined grayscale distribution. Then, the correlation model obtained by the fitting equation above is used to calculate and predict the distribution of segregated carbon content in the 70 steel continuous casting billet with a cross-sectional size of 150 mm × 150 mm. The results are shown in Figure 2. Figure 4 shown.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit the technical solution. Those skilled in the art should understand that those modifications or equivalent substitutions of the technical solution of the present invention that do not depart from the purpose and scope of the technical solution should be included in the scope of the claims of the present invention.

Claims

1. A brightness combination optimization method for predicting the distribution of alloy segregation element content, characterized in that: The following steps are involved: (1) Display of sample cross-section microstructure: According to the process requirements, the cross-section of the alloy microstructure sample is acid-etched to display the morphology of the cross-section of the alloy microstructure sample; (2) Obtaining sample cross-section microstructure images: Using photographic equipment to take photos of the sample cross-section under different light source brightness conditions, obtain low-magnification microstructure morphology images of the cross-section of the continuous casting billet sample after acid etching; (3) Analysis of image quality characteristics and grayscale distribution of tissues at different brightness: Select the complete area including the starting and ending positions of segregation for electron probe or chemical analysis to obtain the one-dimensional or two-dimensional actual content distribution of segregation elements, analyze the image quality characteristics and grayscale distribution of low-magnification tissues through the image information entropy of the cross-sectional tissue, and / or combine the fitting results of the grayscale distribution and the actual content distribution of the segregation elements to determine the optimal brightness M for predicting the content distribution of segregation elements based on the overall tissue image; then use the same method to analyze the area with high content of segregation elements to obtain the optimal brightness L of the corresponding area; then use the same method to analyze the area with low content of segregation elements to obtain the optimal brightness H of the corresponding area; (4) Grayscale combination and model construction of images with different brightness: sort the segregation element content from high to low, match and sort the grayscale values ​​of the three different brightness levels of L, M, and H with the segregation element content, and then divide the segregation element content from high to low into three intervals: high content area, medium content area, and low content area, and obtain the dividing element content C1 of the high content area and the medium content area and the dividing element content C2 of the medium content area and the low content area; then for the high content area, take the grayscale value under L brightness; for the medium content area, take the grayscale value under M brightness; for the low content area, take the grayscale value under H brightness, and thus obtain the grayscale critical value G1 that distinguishes the high content area and the medium content area and the grayscale critical value G2 that distinguishes the medium content area and the low content area; based on the above values, obtain the grayscale combination distribution that matches the segregation element content distribution after combination; based on the grayscale combination distribution and the segregation element content distribution, construct a correlation model between grayscale and element content by fitting equations. (5) Model application: First, for other unmeasured areas or new samples of the same type, the image grayscale values ​​corresponding to the three different brightness levels of L, M, and H are obtained and sorted from small to large. The sorted grayscale values ​​are divided into three intervals using G1 and G2 as the dividing values. Then, the grayscale value less than G1 is taken at L brightness, the grayscale value between G1 and G2 is taken at M brightness, and the grayscale value greater than G2 is taken at H brightness, thereby obtaining the combined grayscale distribution. Then, the correlation model obtained in the previous step is used to calculate and predict the distribution of segregation element content in other unmeasured areas or new samples of the same type.

2. The brightness combination optimization method for predicting the content distribution of alloy segregation elements according to claim 1 is characterized in that: The acid etching treatment is based on the combined action of ultrasound and acid etching reaction, and the vibration frequency of the ultrasonic acid etching device is set to 20-50KHz, the power density is set to 0.5-1.0W / cm2, and the acid etching time is set to 5-20min.

3. According to the brightness combination optimization method for predicting the distribution of alloy segregation element content according to claim 1 or 2, the acid etching liquid for acid etching treatment is a hydrochloric acid alcohol aqueous solution with a volume ratio of 1.0: (0.1-0.5): (0.5-2.0), or a hydrochloric acid sulfuric acid alcohol aqueous solution with a volume ratio of 3.0: (0.5-1.0): (0.1-0.5): (0-4.0); or a hydrochloric acid nitric acid alcohol solution with a volume ratio of (0-5.0): (0.5-1.0): (5.0-15.0), and the heating temperature of the solution in the acid etching tank is 20-70°C.

4. The brightness combination optimization method for predicting the content distribution of alloy segregation elements according to claim 1 or 2, characterized in that: In step (2), except for the brightness of the light source, other parameters of the image acquisition process including the distance between the sample and the light source and the image acquisition range are the same. The different color or grayscale of the image reflects the illumination intensity of different colors of light in the light source.

5. The brightness combination optimization method for predicting the content distribution of alloy segregation elements according to claim 1 or 2, characterized in that: The image information entropy of the cross-sectional structure under different brightness conditions and the image grayscale distribution of the segregation element content areas at different positions are obtained, and compared with the actual ingot segregation element distribution; the image information entropy is calculated by formula (1): Where l is the color level of the pixel in the tissue image; L is the maximum color level, L = 255; P(l) is the proportion of pixels with color level l in the tissue morphology image, and E is the image information entropy. The grayscale distribution of the image is compared with the actual segregation element content. The determination coefficient R of the result is calculated by fitting the equation 2 To judge, R 2 The larger the value, the better the fitting effect. 2 The smaller it is, the worse the fitting effect is; 2 Calculated by formula (2): Where: u i is the measured value; is the model prediction value; is the average of the measured values.

6. The brightness combination optimization method for predicting the content distribution of alloy segregation elements according to claim 5, characterized in that: The fitting equation is fitted and constructed using formula (3): Where: C i is the content of segregation elements at a certain position of the alloy calculated by the model; is the mean value of alloy segregation element content; n is the correction coefficient of the mean value of ingot segregation element content; k0 is the equilibrium distribution coefficient of the solute of the segregation element; a is the correction coefficient of the equilibrium distribution coefficient of the solute of the segregation element; g i The gray value of the tissue image at a certain position point of the billet; is the grayscale mean in the grayscale image of the ingot structure; b is the correction coefficient of the ratio parameter of the grayscale value at a certain position in the grayscale image of the ingot structure to the grayscale mean.

Images