Tin-based babbitt metal metallographic structure quantitative analysis method based on Image J and Cellpose

Abstract

The invention provides a tin-based babbitt alloy metallographic structure quantitative analysis method based on Image J and Cellpose, which comprises the following steps: alloy metallographic preparation and image acquisition are carried out, Image J is used to carry out metallographic image preprocessing and image classification is carried out according to phase characteristics, Image J and Cellpose classification are combined to carry out alloy structure beta '-SnSb and e-Cu6Sn5 phase segmentation and characteristic parameter measurement on a picture, and a tin-based babbitt alloy metallographic structure quantitative analysis result is obtained. The method comprises the following steps of: obtaining geometrical characteristic parameters of a characteristic region phase, performing classifier iterative training by applying a Trainable Weka Segmentation plug-in of Image J software to realize segmentation of picture characteristics close to two phases, obtaining geometrical parameters of an e-Cu6Sn5 phase characteristic region, obtaining phase distribution uniformity parameters by adopting a grid area method and a Voronoi diagram method, and realizing quantitative expression of alloy metallographic structure uniformity. And finally, the number, the phase size, the phase fraction, the phase shape parameter and the phase distribution uniformity of the tin-based babbitt metal phases are obtained. According to the method, a complex programming process with an access threshold is avoided to carry out image segmentation processing, man-made interference is eliminated, geometric feature quantitative parameters of the alloy metallographic structure are obtained more intelligently, quickly and automatically, and the measurement efficiency and accuracy are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of alloy metallographic structure analysis, relates to the measurement of phase fraction, phase size, phase shape and distribution uniformity of an alloy, and specifically relates to a metallographic structure quantitative analysis method of a tin-based babbitt alloy based on Image J and Cellpose. Background Art

[0002] Tin-based babbitt alloy (Sn-Sb-Cu) is widely used as a metal material in sliding bearings due to its remarkable tribological properties (such as wear resistance, embedding and seizure resistance). It is applied to fields involving sliding bearings such as steam turbines, engines, ball mills, etc. Tin-based babbitt alloys usually contain about 3-8 wt.% copper and 5-12 wt.% antimony, and the microstructure of the alloy varies depending on its composition. When the alloy composition contains about 0.5–8 wt.% Cu and less than 8 wt.% Sb, the alloy structure consists of a solid solution soft matrix α phase (Sb dissolved in Sn solid solution) and a needle-shaped or star-shaped hard phase e-Cu6Sn5 phase and a round fine-grained SbSn phase distributed in the matrix phase. The proportion of the η phase component increases with the increase of the copper content. The metallographic structure mainly shows a black α matrix phase and an e-Cu6Sn5 phase. When the alloy composition contains about 0.5–8% Cu and more than 8 wt.% Sb, the alloy structure is composed of α matrix phase, cubic β'-SnSb, and e-Cu6Sn5 distributed in the matrix. The structure of the ingot alloy directly affects the quality of the product, and also affects its hardness, strength and other properties. When it is used in bearings, the alloy structure is also related to the compressive strength, wear resistance, fatigue resistance and service life of the bearing. Obtaining the characteristic parameters of each phase in the tin-based babbitt alloy structure, such as the size and phase fraction of the phase, will be more conducive to quantitatively studying the relationship between the alloy structure and performance. Regarding the detection and analysis method of tin-based babbitt alloy structure, the standard "Tin-based Bearing Alloy Metallographic Inspection" CB1156-92 stipulates the metallographic sampling and preparation of bearings, the alloy structure evaluation method and the grade diagram. This standard evaluates the alloy structure grade based on the size of the phase of the tin-based babbitt alloy, such as the side length of the β'-SnSb phase and the length of the e-Cu6Sn5 phase. Usually, in order to study the microstructure and organizational characteristics of alloys, manual measurement or commercial metallographic analysis software such as Photoshop, Image-pro plus, MIPAR, etc. is used, or image segmentation and recognition are performed through programming to semi-automatically or fully automatically measure the alloy grain size and quantity statistics, grain morphology, grain boundary and phase size, phase fraction and other characteristic parameters to achieve alloy organization analysis and establish the relationship between organization and performance.

[0003] Image J is a free open source image analysis software developed by the National Institutes of Health, which is widely used in the processing and analysis of images in the biomedical field. The software can calculate a series of geometric features of the analysis object in the selected area, including length, angle, perimeter, area, major axis, minor axis, roundness, best ellipse fitting, minimum circumscribed rectangle fitting, and centroid coordinates, and can process images in batches. The Chinese patent with publication number CN110570391A discloses an image analysis method for spray freezing coating effect based on Image J, and uses Image J software to process and analyze the spray freezing coating photos taken by a high-definition camera to obtain the surface coating effect of inert carrier particles. Chinese patent CN115222716A discloses a method for crack morphology recognition based on Image J software, using ImageJ to adjust the threshold and color threshold, find the main strain concentration area (i.e. the area where the crack is located), realize the measurement of the overall morphology and length of the crack, and improve the accuracy of quantitative detection of cracks. Chinese patent CN113758840A discloses a method for measuring the particle size and area of ​​soot particles based on Image J. The invention uses Image J to adjust different thresholds to select the soot area, and finally performs area measurement, which greatly reduces the influence of the image background color on the measurement of the soot particle area, and batches multiple pictures. Chinese patent CN112967176A discloses a method for analyzing plant coverage using Image J and Photoshop, which uses Image J software and Photoshop software to jointly perform data analysis, correct the image geometric distortion caused by the tilted shooting of the coverage image, so that it can meet the requirements of instant shooting and measurement, thereby achieving the purpose of quickly quantifying the change of surface plant coverage. In addition, there are also cases where Image J is applied to the analysis of metal material organization, such as the document "Analysis of cemented carbide grain size based on Image J". Using Image J to measure the grain size of the alloy can effectively segment the adhesion WC-Co cemented carbide grains, and can quickly extract the edges, thereby reducing the misjudgment of the number of grains, reducing manual intervention, achieving automation, and improving the accuracy of measurement.

[0004] However, there are few reports on quantitative analysis methods for tin-based babbitt alloy structure, especially because the β'-SnSb phase and e-Cu6Sn5 phase in the SnSb11Cu6 alloy structure are overlapped due to the solidification characteristics, and the metallographic structure photos show that the two phases are grayish white with similar color characteristics. When using ImageJ to analyze its structure, the two phases cannot be effectively identified, resulting in measurement deviation. Fortunately, ImageJ supports a large number of plug-ins and scripts. Users can install and use different plug-ins according to their needs, or write scripts to expand the functions of Image J. Cellpose software can realize automatic identification and segmentation of irregularly shaped, overlapping and adherent cells. The software was developed by Carsen Stringer, Timothee Lee, David VanValen, Maximilian Haeussler, Marius Pachitariu, etc. Its source code and documentation can be found at https: / / github.com / MouseLand / cellpose. Cellpose is based on deep learning algorithms, especially convolutional neural networks (CNN), which can automatically identify and segment cells, cell nuclei and other cell structures, provide a user interface (GUI), and can save the segmented and identified feature area information into a format recognizable by Image J and import it into ImageJ software for analysis and processing. Based on the above tools, the present invention uses the open source and free Image J and Cellpose software in combination according to the organizational characteristics of tin-based babbitt alloys, and innovatively performs measurement and analysis of the metallographic structure of tin-based babbitt alloys. Through image preprocessing, threshold adjustment binarization, β'-SnSb phase recognition, model training and automatic recognition and segmentation, the segmentation of the overlapping phases of the alloy structure is achieved, the geometric characteristic parameters of the characteristic area phase are obtained, and the automatic and semi-automatic measurement and quantitative analysis of the phase fraction, phase size and distribution uniformity are achieved, human measurement errors are eliminated, complex programming processes are avoided, and measurement efficiency and accuracy are improved. Summary of the invention

[0005] The purpose of the present invention is to solve the problems of difficult segmentation of overlapping phases of alloy in tin-based babbitt alloy and low measurement efficiency, lack of quantitative indicators and low accuracy of phase structure characteristic parameters. Image J and Cellpose software are used together to measure and analyze the metallographic structure of tin-based babbitt alloy. Image preprocessing, threshold adjustment, binarization, β'-SnSb phase recognition, image deep learning model training and automatic recognition and segmentation are proposed to achieve segmentation of overlapping phases of alloy structure, obtain geometric characteristic parameters of characteristic region phases, propose a method for representing the uniformity of tin-based babbitt alloy structure, realize automatic and semi-automatic measurement and quantitative analysis of alloy phase geometric parameters, phase fraction and distribution uniformity, quantitatively express the metallographic structure of tin-based babbitt alloy, eliminate human measurement errors, avoid complex programming processes, and improve measurement efficiency and accuracy.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme: The quantitative analysis method of the metallographic structure of tin-based babbitt alloy based on Image J and Cellpose software includes the following steps: S1. Metallographic preparation of alloy samples: After grinding the tin-based babbitt alloy with sandpaper, it is polished with alumina polishing liquid with a particle size of 3 μm and 0.25 μm in sequence, and finally chemically mechanically polished or corroded with nitric acid alcohol corrosive liquid to reveal the β'-SnSb phase, e-Cu6Sn5 phase and a matrix phase of the alloy; S2. Collection of metallographic images: The prepared metallographic samples are imaged using an inverted or upright metallographic microscope to obtain clear metallographic images; S3. Image preprocessing and classification: The metallographic images obtained in step S2 are batch-processed into 8-bit grayscale images, and the brightness and contrast are adjusted using Image J software to obtain preprocessed images; the preprocessed images are identified and classified according to the alloy composition and the types of phases composed of the metallographic images. The alloy metallographic images composed of needle-shaped e-Cu6Sn5 phases and black a matrix phases are classified into the first category; the alloy metallographic images composed of nearly square β'-SnSb phases and black a matrix phases are classified into the second category; the alloy metallographic images composed of β'-SnSb phases, e-Cu6Sn5 and a matrix phases are classified into the third category; S4. Measurement of metallographic structure characteristic parameters: First, use ImageJ software to set the scale, and then select the following corresponding steps to process and measure according to the metallographic image classification results in step S3 to obtain the structure characteristic parameters: S4.1 For the metallographic image of the first type of alloy composed of e-Cu6Sn5 phase and a matrix phase, use Image J software to adjust the threshold to obtain a binary image, use a manual measurement tool to select at least 20 feature positions to measure the length of the e-Cu6Sn5 phase to obtain the phase size; measure the area ratio of the e-Cu6Sn5 phase in the field of view to obtain the phase fraction of the alloy e-Cu6Sn5 phase; use the grid area method to measure and obtain the distribution uniformity parameters of the e-Cu6Sn5 phase; S4.2 For the metallographic image of the second type of alloy composed of β'-SnSb phase and a matrix phase, use Image J software to adjust the threshold to obtain a binary image, and perform denoising, closing, filling holes or watershed algorithm segmentation preprocessing to obtain a feature image; then use Analyze Particles software to analyze the particles of the β'-SnSb phase in the field of view to obtain the phase number, phase size parameters, phase fraction and phase shape parameters; use the Voronoi diagram method to obtain the β'-SnSb phase distribution uniformity parameters; S4.3 For the metallographic image of the third type of alloy composed of β'-SnSb phase, e-Cu6Sn5 and a matrix phase, first use Cellpose software to identify and segment the β'-SnSb phase and train the model, then use the new model obtained by training to identify and segment the β'-SnSb phase on the new image, and obtain the β'-SnSb phase characteristic image after image operation on the identified and segmented image, and then process it according to the steps of step S4.2 to obtain the phase number, phase size parameters, phase fraction and phase shape parameters of β'-SnSb; S5. Data statistical analysis: The phase number, phase size, phase fraction, phase shape parameter and phase distribution uniformity parameter obtained in step S4 are statistically analyzed to obtain the quantified average value of the alloy phase size, average value of the phase fraction and phase distribution uniformity.

[0007] Furthermore, the chemical mechanical polishing described in step S1 is to polish the tin-based babbitt alloy for 30-120 seconds using an alkaline silica suspension polishing liquid before the metallographic etching of the tin-based babbitt alloy, and directly obtain a metallographic image consisting of the e-Cu6Sn5 phase and the a matrix phase after mechanical chemical polishing and before etching of the alloy consisting of three phases: β'-SnSb phase, e-Cu6Sn5 and a matrix phase.

[0008] Furthermore, in step S3, the method of batch preprocessing images using Image J software is to select File-Image-Image sequence to batch import images as a stacked image set Stack arranged in sequence, perform 8-bit grayscale image conversion, brightness and contrast adjustment on the images in the Stack, and finally use File-Save as-Imagesequence to batch export the preprocessed images.

[0009] Furthermore, the grid area method described in step S4.1 is used to determine and obtain the uniformity parameter of the e-Cu6Sn5 phase distribution, which is to use the ImageJ software to select Image-Stacks-Tool-Montage to stack in sequence to evenly divide the image into grid images of no less than 2´2 grids, and form a stacked image set Stack, and then measure the phase area ratio of each grid image in the Stack to obtain the phase fraction.

[0010] Furthermore, the grid area method described in step S4.1 is used to measure and obtain the e-Cu6Sn5 phase distribution uniformity parameter, which is expressed as: U e-Cu6Sn5 =100*(1-(F SD / F Mean ) Among them, U e-Cu6Sn5 is the uniformity of e-Cu6Sn5 phase distribution, F SD is the standard deviation of the e-Cu6Sn5 phase fraction in the cut grid image, F Mean It is the average value of the e-Cu6Sn5 phase fraction in the cut grid image.

[0011] Furthermore, the Voronoi diagram method described in step S4.2 is used to obtain the β'-SnSb phase distribution uniformity parameter, which is to use ImageJ software to select Process-Bianary-Voronoi in sequence on the binary feature image to obtain the SnSb phase distribution Voronoi diagram, and then use Analyze Particles to obtain the Voronoi diagram area ratio of each phase.

[0012] Furthermore, the uniformity of the β'-SnSb phase distribution obtained by the Voronoi diagram measurement described in step S4.2 is expressed as: U β'-SnSb =100*(1-(A SD / A Mean ))% Among them, U β′-SnSb is the distribution uniformity of β'-SnSb phase, A SD is the standard deviation of the area ratio of the Voronoi diagram of the β'-SnSb phase, AMean is the average value of the area ratio of the Voronoi diagram of the β'-SnSb phase.

[0013] Furthermore, the use of Cellpose software to identify, segment and train the model of the β'-SnSb phase as described in step S4.3 is to first put no less than 10 pictures of the same multiple into the same folder, use Cellpose software to open one of the pictures and use the software's own pre-trained model to automatically identify and segment, then manually delete or correct the misidentified β'-SnSb feature area or manually add and mark the unidentified feature area, and then train a new model, repeat the above steps three or more times to obtain a special model for β'-SnSb phase identification and segmentation, and use the model to automatically identify and segment new pictures.

[0014] Furthermore, the image operation described in step S4.3 is performed by saving the masked image identified and segmented by the Cellpose software, and then using Image calculate of the Image J software to subtract the ROI area from its contour image to obtain the β'-SnSb phase characteristic image.

[0015] Furthermore, for the first type of metallographic image composed of the e-Cu6Sn5 phase and the a matrix phase divided in step S3, the Trainable Weka Segmentation plug-in in the Image J software is used to classify the two phases. The two-phase segmentation is achieved through iterative classifier training to obtain the characteristic area image of the e-Cu6Sn5 phase, and then the phase size, phase fraction, and phase distribution uniformity parameters are obtained according to the method of step S4.1.

[0016] The method of the present invention has the following beneficial effects: 1. The present invention uses Image J and Cellpose software in combination, based on a deep learning image segmentation algorithm, and through model training, realizes automatic recognition and segmentation processing of overlapping and similar β'-SnSb and e-Cu6Sn5 phases in the alloy structure, and accurately obtains the geometric characteristic parameters of the β'-SnSb phase.

[0017] 2. The present invention adopts the grid area method and Voronoi diagram method based on Image J software to obtain the phase distribution uniformity. This method can quantitatively represent the organizational uniformity of the tin-based babbitt alloy.

[0018] 3. The present invention aims at an alloy composed of three phases: β'-SnSb phase, e-Cu6Sn5 and a matrix phase. Metallographic images are obtained before and after chemical mechanical polishing and corrosion, and the geometric characteristic parameters of e-Cu6Sn5 and β'-SnSb phases are obtained based on Image J and Cellpose step-by-step method, thereby avoiding mutual interference between the two phases during measurement and increasing measurement accuracy.

[0019] 4. The Image J and Cellpose software used in the present invention are open source and free software, which does not require the cost of purchasing image processing software. At the same time, it avoids the complex and entry-level programming process for image processing, and obtains the geometric characteristic quantitative parameters of the alloy phase more intelligently, quickly and automatically, eliminates human interference, and increases measurement accuracy and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The steps for manually measuring the length of the e-Cu6Sn5 phase; Figure 2 The process of obtaining the uniformity of e-Cu6Sn5 phase distribution by the grid area method; Figure 3 It is the process of measuring geometric characteristic parameters of alloy β'-SnSb phase; Figure 4 The process of obtaining the uniformity of β'-SnSb phase distribution by Voronoi diagram method; Figure 5 The steps of using Cellpose software to identify, segment and train the model of β'-SnSb phase; Figure 6 The steps of obtaining the uniformity of β'-SnSb phase distribution by image operation and Voronoi diagram method; Figure 7 The steps for obtaining the characteristic regions of the e-Cu6Sn5 phase using the Trainable Weka Segmentation plug-in in ImageJ software; Figure 8 The difference in the uniformity of the metallographic structure of tin-based babbitt alloy; Fig. 9 The difference in the shape of the β'-SnSb phase in the metallographic structure of tin-based babbitt alloy. DETAILED DESCRIPTION

[0021] In order to better understand the content of the present invention, the present invention will be further described below in conjunction with specific embodiments and drawings. The following implementation process is based on the technology of the present invention and provides detailed implementation methods and operating steps, but the protection scope of the present invention is not limited to the following implementation process. Example 1

[0022] A quantitative analysis method of the metallographic structure of SnSb8Cu4 alloy based on Image J software, the specific steps are as follows: S1. Metallographic preparation of alloy samples: After the tin-based babbitt alloy SnSb8Sb4 was ground with sandpaper, it was polished with alumina polishing liquid with a particle size of 3 μm and 0.25 μm in sequence, and then chemically mechanically polished with alkaline silica suspension polishing liquid for 30 seconds, and finally corroded with 4% nitric acid plus alcohol etching solution to reveal the alloy structure.

[0023] S2. Collection of metallographic images: Use an upright metallographic microscope to collect images of the prepared metallographic samples to obtain clear metallographic images.

[0024] S3. Image preprocessing and classification: The images obtained in step S2 are batch converted to 8-bit grayscale images, and the brightness and contrast are adjusted using Image J to obtain preprocessed images. The preprocessed images are identified and classified according to the types of phases composed of the alloy, and the alloy is composed of needle-shaped e-Cu6Sn5 phases and black a matrix phases.

[0025] S4. Measurement of metallographic structure characteristic parameters: For the metallographic image of the alloy composed of e-Cu6Sn5 phase and a matrix phase, first use ImageJ to set the scale, then adjust the threshold to obtain a binary image, use the straight line selection tool to manually measure and select at least 20 e-Cu6Sn5 phase characteristic positions, and add them to the ROI Manager, then measure the phase length to obtain the phase size, such as Figure 1 As shown; the area ratio of the e-Cu6Sn5 phase in the field of view of the picture is measured to obtain the phase fraction of the alloy e-Cu6Sn5 phase.

[0026] The grid area method was used to measure and obtain the uniformity parameters of the e-Cu6Sn5 phase distribution. The specific steps are: Use the ImageJ software to select Image-Stacks-Tool-Montage to stack to evenly divide the image into 3'3 grid grids and form a stacked image set Stack, and then measure the phase area ratio of the 9 grid images in the Stack to obtain the phase fraction, such as Figure 2 The uniformity of e-Cu6Sn5 phase distribution (%) obtained by the grid area method is expressed as: U e-Cu6Sn5 =100*(1-(F SD / F Mean ) Among them U e-Cu6Sn5 is the uniformity of e-Cu6Sn5 phase distribution, F SD is the standard deviation of the e-Cu6Sn5 phase fraction in the cut grid image, FMean It is the average value of the e-Cu6Sn5 phase fraction in the cut grid image.

[0027] S5. Data statistical analysis: The phase size, phase fraction and phase distribution uniformity parameters obtained in step S4 are statistically analyzed to obtain the quantitative data of the alloy structure such as the average value and standard deviation of the corresponding parameters, as shown in Table 1. The length of the alloy e-Cu6Sn5 phase is 84.81 mm, the phase fraction is 15.3%, and the phase structure uniformity is 93.57%. Example 2

[0028] A quantitative analysis method of the metallographic structure of SnSb10.5Cu0.02Ni0.02 alloy based on Image J software, the specific steps are as follows: S1. Metallographic preparation of alloy samples: After the tin-based babbitt alloy SnSb10.5Cu0.02Ni0.02 was ground with sandpaper, it was polished with alumina polishing liquid with a particle size of 3 μm and 0.25 μm in sequence, and then chemically mechanically polished with alkaline silica suspension polishing liquid for 30 seconds, and finally corroded with 4% nitric acid plus alcohol etching solution to reveal the alloy structure.

[0029] S2. Collection of metallographic images: Use an upright metallographic microscope to collect images of the prepared metallographic samples to obtain clear metallographic images.

[0030] S3. Image preprocessing and classification: The images obtained in step S2 were batch converted to 8-bit grayscale images, and the brightness and contrast were adjusted using Image J to obtain preprocessed images. The preprocessed images were identified and classified according to the types of phases composed of the alloy. The alloy structure consisted of a nearly square β'-SnSb phase and a black a matrix phase.

[0031] S4. Measurement of metallographic structure characteristic parameters: For the metallographic image of the alloy composed of β'-SnSb phase and a matrix phase, first use Image J to set the scale, then adjust the threshold to obtain a binary image, and perform denoising, closing, filling holes or watershed algorithm segmentation preprocessing to obtain a characteristic image; then use Analyze Particles to analyze the particles of the β'-SnSb phase in the field of view to obtain the phase number, phase size parameters, phase fraction and phase shape parameters, such as Figure 3 The phase shape parameter is the ratio of the area enclosed by the outer contour of the β'-SnSb phase to the area enclosed by the convex hull of the outer contour of the β'-SnSb phase. The larger the value, the more likely the β'-SnSb phase is to nucleate alone without obvious protrusions. The smaller the value, the more likely the phase is to aggregate or grow differently, resulting in more convex hulls or branches in the β'-SnSb phase in the metallographic image.

[0032] The Voronoi diagram method is used to obtain the uniformity parameters of the β'-SnSb phase distribution. Specifically, the binary feature images are processed using ImageJ software and Process-Bianary-Voronoi is selected in sequence to obtain the SnSb phase distribution Voronoi diagram, and then the Analyze Particles is used to obtain the Voronoi diagram area of ​​each phase, such as Figure 4 The distribution uniformity (%) of β'-SnSb phase obtained by Voronoi diagram method is expressed as: U β'-SnSb =100*(1-(A SD / A Mean ) Among them U β′-SnSb is the distribution uniformity of β'-SnSb phase, A SD is the standard deviation of the area ratio of the Voronoi diagram of the β'-SnSb phase, A Mean is the average value of the area ratio of the Voronoi diagram of the β'-SnSb phase.

[0033] S5. Data statistical analysis: The number of phases, phase size, phase fraction, phase shape parameters and phase distribution uniformity parameters obtained in step S4 are statistically analyzed to obtain quantitative data of alloy structure such as average value and standard deviation, as shown in Table 1. The size of the β'-SnSb phase refers to the side length of the phase, and the phase is approximated as a square, and its side length is the measured phase perimeter divided by 4. Example 3

[0034] A SnSb11Cu6 alloy metallographic structure quantitative analysis method based on Image J software and Cellpose software, the specific steps are as follows: S1. Metallographic preparation of alloy samples: After grinding the tin-based babbitt alloy SnSb11Cu6 with sandpaper, it was polished with alumina polishing liquid with particle sizes of 3 μm and 0.25 μm in sequence, and then chemically mechanically polished with alkaline silica suspension polishing liquid for 30 seconds, and finally corroded with 4% nitric acid plus alcohol etching solution to reveal the alloy structure.

[0035] S2. Collection of metallographic images: Use an upright metallographic microscope to collect images of the prepared metallographic samples to obtain clear metallographic images.

[0036] S3. Image preprocessing and classification: The images obtained in step S2 are batch converted to 8-bit grayscale images, and the brightness and contrast are adjusted using Image J to obtain preprocessed images. The preprocessed images are identified and classified according to the types of phases composed of the alloy. The alloy consists of three phases: β'-SnSb phase, e-Cu6Sn5 and a matrix phase.

[0037] S4. Measurement of metallographic structure characteristic parameters: For the metallographic image of the alloy composed of β'-SnSb phase, e-Cu6Sn5 and a matrix phase, the β'-SnSb phase is first identified and segmented using the Cellpose software. First, no less than 10 pictures of the same magnification are placed in the same folder, one of the pictures is opened with Cellpose and the software's own pre-trained model is used to automatically identify and segment the β'-SnSb phase and e-Cu6Sn5. After the initial segmentation is completed, the incorrectly identified β'-SnSb feature area (Mask) is manually deleted or corrected or the unidentified Mask is manually marked and added, and then a new model is trained. Repeat the above steps more than three times to obtain a dedicated model for β'-SnSb phase identification and segmentation, such as Figure 5 As shown. Use a dedicated new model to automatically identify and segment new images, save the images containing masks that have been identified and segmented by Cellpose software, and use Image J to select Process-Image calculate to subtract the ROI area of ​​the Mask image from its contour image to obtain the β'-SnSb phase feature image, so as to avoid the adhesion of adjacent masks after binarization. Then use Analyze Particles to perform particle analysis on the β'-SnSb phase in the field of view to obtain the number of phases, phase size parameters, phase fraction and phase shape parameters. The Voronoi diagram method is used to obtain the uniformity parameters of the β'-SnSb phase distribution. Specifically, use Image J software to select Process-Bianary-Voronoi in turn for the binary feature image to obtain the SnSb phase distribution Voronoi diagram, and then use Analyze Particles to obtain the Voronoi diagram area of ​​each phase, as shown below. Figure 6 shown.

[0038] The distribution uniformity (%) of β'-SnSb phase obtained by Voronoi diagram method is expressed as: U β'-SnSb =100*(1-(A SD / A Mean ) Among them U β'-SnSb is the distribution uniformity of β'-SnSb phase, A SD is the standard deviation of the area ratio of the Voronoi diagram of the β'-SnSb phase, A Mean is the average value of the area ratio of the Voronoi diagram of the β'-SnSb phase.

[0039] S5. Data statistical analysis: The number of phases, phase size, phase fraction, phase shape parameter and phase distribution uniformity parameter obtained in step S4 are statistically analyzed to obtain quantitative alloy structure data such as average value and standard deviation, as shown in Table 1. Example 4

[0040] A quantitative analysis method of the metallographic structure of SnSb11Cu6 alloy based on Image J software, the specific steps are as follows: S1. Metallographic preparation of alloy samples: After the tin-based babbitt alloy SnSb11Cu6 was ground with sandpaper, it was polished with alumina polishing liquid with particle sizes of 3 μm and 0.25 μm in sequence, and finally chemically mechanically polished with alkaline silica suspension polishing liquid for 100 seconds to make the alloy structure visible.

[0041] S2. Collection of metallographic images: Use an upright metallographic microscope to collect images of the prepared metallographic samples to obtain clear metallographic images.

[0042] S3. Image preprocessing and classification: The images obtained in step S2 are batch converted to 8-bit grayscale images, and brightness and contrast are adjusted using Image J to obtain preprocessed images. The preprocessed images are identified and classified according to the types of phases composed of the alloy, and the alloy is composed of e-Cu6Sn5 and a matrix phase.

[0043] S4. Measurement of metallographic structure characteristic parameters: The metallographic image composed of the e-Cu6Sn5 phase and the a matrix phase obtained after mechanical chemical polishing has similar image features and cannot be distinguished by direct binarization. Therefore, the Trainable Weka Segmentation plug-in in the ImageJ software is used to manually select the e-Cu6Sn5 phase and the a matrix phase, and the classifier is trained. After multiple iterations, the e-Cu6Sn5 phase and the a matrix phase image are segmented, and finally the e-Cu6Sn5 phase characteristic area image is obtained by Createresult, as shown in the figure. Figure 7 Then, the characteristic parameters of the e-Cu6Sn5 phase are obtained according to steps S4 and S5 described in Example 1, as shown in Table 1. Example 5

[0044] A quantitative analysis method for the metallographic structure of tin-based babbitt alloy based on Image J software and Cellpose software. This embodiment mainly demonstrates the measurement method of the alloy metallographic structure uniformity difference, and selects the alloy metallographic structure photos of different positions (position A, position B) of SnSb12Cu6 alloy ingot for comparison. The specific steps are as follows: S1. Metallographic preparation of alloy samples: After grinding the tin-based babbitt alloy SnSb12Cu6 with sandpaper, it was polished with alumina polishing liquid with particle sizes of 3 μm and 0.25 μm in sequence, and then chemically mechanically polished with alkaline silica suspension polishing liquid for 30 seconds, and finally corroded with 4% nitric acid plus alcohol etching solution to reveal the alloy structure.

[0045] S2. Collection of metallographic images: The prepared metallographic samples are collected using an inverted or upright metallographic microscope. Two different positions, the edge and the middle of the alloy ingot, are selected during collection to obtain clear metallographic images.

[0046] S3. Image preprocessing and classification: The images obtained in step S2 are batch converted to 8-bit grayscale images, and the brightness and contrast are adjusted using Image J to obtain preprocessed images. The preprocessed images are identified and classified according to the types of phases composed of the alloy. The alloy consists of three phases: β'-SnSb phase, e-Cu6Sn5 and a matrix phase.

[0047] S4. Measurement of metallographic structure characteristic parameters: For the metallographic images of the alloy composed of β'-SnSb phase, e-Cu6Sn5 and a matrix phase, first use the dedicated model of Cellpose software for β'-SnSb phase identification and segmentation to automatically identify and segment the two images. After saving the images with masks identified and segmented by Cellpose software, use Image J to select Process-Image calculate to subtract the ROI area of ​​the mask image from its contour image to obtain the β'-SnSb phase characteristic image, so as to avoid the adhesion of adjacent masks after binarization. The Voronoi diagram method is used to obtain the uniformity parameters of the β'-SnSb phase distribution. Specifically, the binarized characteristic images are selected in turn by ImageJ software Process-Bianary-Voronoi to obtain the SnSb phase distribution Voronoi diagram, and then Analyze Particles is used to obtain the Voronoi diagram area ratio of each phase in the two metallographic images, as shown in the figure. Figure 8 shown.

[0048] The distribution uniformity (%) of β'-SnSb phase obtained by Voronoi diagram method is expressed as: U β'-SnSb =100*(1-(A SD / A Mean ) Among them U β′-SnSb is the distribution uniformity of β'-SnSb phase, A SD is the standard deviation of the area ratio of the Voronoi diagram of the β'-SnSb phase, AMean is the average value of the area ratio of the Voronoi diagram of the β'-SnSb phase.

[0049] S5. Data statistical analysis: The phase distribution uniformity parameters obtained in step S4 are statistically analyzed, as shown in Table 1, and the tissue uniformity of the two images is compared. Example 6

[0050] A quantitative analysis method of the metallographic structure of tin-based babbitt alloy based on Image J software and Cellpose software. This embodiment mainly demonstrates the measurement of the shape difference of the alloy β'-SnSb phase, and selects the alloy metallographic structure photos of two SnSb11Cu6 alloy ingots under different preparation conditions (C condition and D condition) for comparison. The specific steps are as follows: S1. Metallographic preparation of alloy samples: After grinding the tin-based babbitt alloy SnSb11Cu6 with sandpaper, it was polished with alumina polishing liquid with particle sizes of 3 μm and 0.25 μm in sequence, and then chemically mechanically polished with alkaline silica suspension polishing liquid for 30 seconds, and finally corroded with 4% nitric acid plus alcohol etching solution to reveal the alloy structure.

[0051] S2. Collection of metallographic images: The prepared metallographic samples are collected using an inverted or upright metallographic microscope. Two different positions, the edge and the middle of the alloy ingot, are selected during collection to obtain clear metallographic images.

[0052] S3. Image preprocessing and classification: The images obtained in step S2 are batch converted to 8-bit grayscale images, and the brightness and contrast are adjusted using Image J to obtain preprocessed images. The preprocessed images are identified and classified according to the types of phases composed of the alloy. The alloy consists of three phases: β'-SnSb phase, e-Cu6Sn5 and a matrix phase.

[0053] S4. Measurement of metallographic structure characteristic parameters: For the metallographic image of the alloy composed of β'-SnSb phase, e-Cu6Sn5 and a matrix phase, the β'-SnSb phase is first identified and segmented using the Cellpose software. The specific steps are as follows: first, put no less than 10 pictures of the same alloy with the same multiple into the same folder, open one of the pictures using Cellpose and use the software's own pre-trained model to automatically identify and segment, so as to achieve the segmentation of the β'-SnSb phase and e-Cu6Sn5. After the initial segmentation is completed, manually delete or correct the misidentified β'-SnSb feature region (ROI) or manually add the unidentified ROI, and then train the new model. Repeat the above steps for more than three times to obtain a special model for β'-SnSb phase identification and segmentation, and use this model to automatically identify and segment the new picture. The other alloy also uses the above steps to perform the special model and automatic identification and segmentation of the β'-SnSb phase identification and segmentation. After saving the masked picture identified and segmented by the Cellpose software, use Image J to select Process-Image Calculate the subtraction of the ROI area of ​​the Mask image and its contour image to obtain the β'-SnSb phase characteristic image, avoiding the adhesion of adjacent masks after binarization. Then use AnalyzeParticles to analyze the particles of the β'-SnSb phase in the field of view and obtain the phase shape parameters, such as Fig. 9 shown.

[0054] S5. Data statistical analysis: The phase shape parameters obtained in step S4 are statistically analyzed to obtain the mean value and standard deviation quantitative data, as shown in Table 1, and the phase shape differences of the two images are compared.

[0055] Table 1 shows the quantitative data of the tin-based babbitt alloy structure obtained in each embodiment .

Claims

1. A quantitative analysis method of the metallographic structure of tin-based babbitt alloy based on Image J and Cellpose, characterized in that: The steps include: S1. Metallographic preparation of alloy samples: After grinding the tin-based babbitt alloy with sandpaper, it is polished with alumina polishing liquid with a particle size of 3 μm and 0.25 μm in sequence, and finally chemically mechanically polished or corroded with nitric acid alcohol corrosive liquid to reveal the β'-SnSb phase, e-Cu6Sn5 phase and a matrix phase of the alloy; S2. Collection of metallographic images: The prepared metallographic samples are imaged using an inverted or upright metallographic microscope to obtain clear metallographic images; S3. Image preprocessing and classification: The metallographic images obtained in step S2 are batch-processed into 8-bit grayscale images, and the brightness and contrast are adjusted using Image J software to obtain preprocessed images; the preprocessed images are identified and classified according to the alloy composition and the types of phases composed of the metallographic images. The alloy metallographic images composed of needle-shaped e-Cu6Sn5 phases and black a matrix phases are classified into the first category; the alloy metallographic images composed of nearly square β'-SnSb phases and black a matrix phases are classified into the second category; the alloy metallographic images composed of β'-SnSb phases, e-Cu6Sn5 and a matrix phases are classified into the third category; S4. Measurement of metallographic structure characteristic parameters: First, use ImageJ software to set the scale, and then select the following corresponding steps to process and measure according to the metallographic image classification results in step S3 to obtain the structure characteristic parameters: S4 .1 For the metallographic image of the first type of alloy composed of e-Cu6Sn5 phase and a matrix phase, use Image J software to adjust the threshold to obtain a binary image, use a manual measurement tool to select at least 20 feature positions to measure the length of the e-Cu6Sn5 phase to obtain the phase size; measure the area ratio of the e-Cu6Sn5 phase in the field of view to obtain the phase fraction of the alloy e-Cu6Sn5 phase; use the grid area method to measure and obtain the distribution uniformity parameters of the e-Cu6Sn5 phase; S4.2 For the metallographic image of the second type of alloy composed of β'-SnSb phase and a matrix phase, use Image J software to adjust the threshold to obtain a binary image, and perform denoising, closing, filling holes or watershed algorithm segmentation preprocessing to obtain a feature image; then use Analyze Particles software to analyze the particles of the β'-SnSb phase in the field of view to obtain the phase number, phase size parameters, phase fraction and phase shape parameters; use the Voronoi diagram method to obtain the β'-SnSb phase distribution uniformity parameters; S4.3 For the metallographic image of the third type of alloy composed of β'-SnSb phase, e-Cu6Sn5 and a matrix phase, first use Cellpose software to identify and segment the β'-SnSb phase and train the model, then use the new model obtained by training to identify and segment the β'-SnSb phase on the new image, and obtain the β'-SnSb phase characteristic image after image operation on the identified and segmented image, and then process it according to the steps of step S4.2 to obtain the phase number, phase size parameters, phase fraction and phase shape parameters of β'-SnSb; S5. Data statistical analysis: The phase number, phase size, phase fraction, phase shape parameter and phase distribution uniformity parameter obtained in step S4 are statistically analyzed to obtain the quantified average value of the alloy phase size, average value of the phase fraction and phase distribution uniformity.

2. The method for quantitative analysis of metallographic structure of tin-based babbitt alloy based on Image J and Cellpose according to claim 1, characterized in that: The chemical mechanical polishing described in step S1 is to polish the tin-based babbitt alloy for 30-120 seconds using an alkaline silica suspension polishing liquid before the metallographic etching of the tin-based babbitt alloy, and directly obtain a metallographic image consisting of the e-Cu6Sn5 phase and the a matrix phase after mechanical chemical polishing and before etching of the alloy consisting of three phases: β'-SnSb phase, e-Cu6Sn5 and a matrix phase.

3. The method for quantitative analysis of metallographic structure of tin-based babbitt alloy based on Image J and Cellpose according to claim 1, characterized in that: The method for batch preprocessing images using Image J software in step S3 is to select File-Image-Image sequence to batch import images as a stacked image set Stack arranged in sequence, perform 8-bit grayscale image conversion, brightness and contrast adjustment on the images in the Stack, and finally use File-Save as-Imagesequence to batch export the preprocessed images.

4. The method for quantitative analysis of metallographic structure of tin-based babbitt alloy based on Image J and Cellpose according to claim 1, characterized in that: The grid area method described in step S4.1 is used to determine the uniformity parameters of the e-Cu6Sn5 phase distribution. The ImageJ software is used to select Image-Stacks-Tool-Montage to stack in sequence to evenly divide the image into grid images of no less than 2´2 squares, and form a stacked image set Stack, and then the phase area ratio of each grid image in the Stack is measured to obtain the phase fraction.

5. A quantitative analysis method of metallographic structure of tin-based babbitt alloy based on Image J and Cellpose according to claim 1 or 4, characterized in that: The grid area method described in step S4.1 is used to measure and obtain the e-Cu6Sn5 phase distribution uniformity parameter, which is expressed as: U e-Cu6Sn5 =100*(1-(F SD / F Mean )) Among them, U e-Cu6Sn5 is the uniformity of e-Cu6Sn5 phase distribution, F SD is the standard deviation of the e-Cu6Sn5 phase fraction in the cut grid image, F Mean It is the average value of the e-Cu6Sn5 phase fraction in the cut grid image.

6. The method for quantitatively analyzing the metallographic structure of tin-based babbitt alloy based on Image J and Cellpose according to claim 1, characterized in that: The Voronoi diagram method described in step S4.2 is used to obtain the β'-SnSb phase distribution uniformity parameter, which is to use the ImageJ software to select Process-Bianary-Voronoi in sequence on the binary feature image to obtain the SnSb phase distribution Voronoi diagram, and then use Analyze Particles to obtain the Voronoi diagram area ratio of each phase.

7. A quantitative analysis method of metallographic structure of tin-based babbitt alloy based on Image J and Cellpose according to claim 1 or 6, characterized in that: The uniformity of the β'-SnSb phase distribution obtained by the Voronoi diagram described in step S4.2 is expressed as: U β'-SnSb =100*(1-(A SD / A Mean ))% Among them, U β′-SnSb is the distribution uniformity of β'-SnSb phase, A SD is the standard deviation of the area ratio of the Voronoi diagram of the β'-SnSb phase, A Mean is the average value of the area ratio of the Voronoi diagram of the β'-SnSb phase.

8. The method for quantitatively analyzing the metallographic structure of tin-based babbitt alloy based on Image J and Cellpose according to claim 1, characterized in that: The use of Cellpose software to identify, segment and train the model of the β'-SnSb phase as described in step S4.3 is to first put no less than 10 pictures of the same multiple into the same folder, use Cellpose software to open one of the pictures and use the software's own pre-trained model to automatically identify and segment, then manually delete or correct the misidentified β'-SnSb feature area or manually add and mark the unidentified feature area, and then train a new model, repeat the above steps three or more times to obtain a special model for β'-SnSb phase identification and segmentation, and use the model to automatically identify and segment new pictures.

9. A method for quantitative analysis of metallographic structure of tin-based babbitt alloy based on Image J and Cellpose according to claim 1 or 8, characterized in that: The image operation described in step S4.3 is performed by saving the masked image identified and segmented by the Cellpose software, and then using the Image calculate of the Image J software to subtract the ROI area from its contour image to obtain the β'-SnSb phase characteristic image.

10. The method for quantitatively analyzing the metallographic structure of tin-based babbitt alloy based on Image J and Cellpose according to claim 1, characterized in that: For the first type of metallographic image composed of e-Cu6Sn5 phase and a matrix phase divided in step S3, the Trainable Weka Segmentation plug-in in Image J software is used to classify the two phases. The two-phase segmentation is achieved through iterative classifier training to obtain the characteristic area image of the e-Cu6Sn5 phase. Then, the phase size, phase fraction, and phase distribution uniformity parameters are obtained according to the method in step S4.1.

Images