Method, system and equipment for intelligently rating grain size of steel and medium

Abstract

The invention relates to the field of steel rating, and provides an intelligent steel grain size rating method, system, equipment and medium, and the method comprises the following steps: obtaining to-be-rated steel and preparing to obtain a grain size sample; performing scanning and view field acquisition on the grain size sample to obtain a metallographic picture; processing the metallographic picture through a preset intelligent rating model to obtain a grain size rating; and testing the accuracy of the grain size rating result to obtain the accuracy, and outputting the grain size rating and the accuracy. According to the method, the intelligent grain size rating model is established by improving the preparation quality of the grain size sample and adopting methods such as convolutional neural network deep learning and the like based on artificial intelligence and image recognition technologies. The automatic metallographic microscope is used for automatically scanning a metallographic sample and automatically collecting a view field, corresponding pictures are reserved and graded according to an average value, and the method has the characteristics of high detection speed, low manual dependence and good repeatability of a detection result.

Description

Technical Field

[0001] The present invention relates to the field of steel material rating, and in particular to a method, system, equipment and medium for intelligent rating of steel material grain size. Background Art

[0002] Compared with the traditional manual inspection and rating method of steel grain size, automatic quantitative inspection has become the mainstream research direction in the industry. Domestic peers have done a lot of research on this. Among them, Cheng Kangna et al., in "Quantitative Analysis of NbC / Ni_3Si-based Composite Coating Based on Digital Image Processing" (Special Casting and Nonferrous Alloys, 2015, 35(7)), used digital image processing methods to process metallographic images, and proposed a method for quantitative analysis of laser cladding metallographic structure with the help of MATLAB and ImageJ software tools to improve the measurement accuracy. Wu Wei et al., in "Quantitative Analysis of Metallographic Structure of TC4 Titanium Alloy Based on Digital Image Processing" (Failure Analysis and Prevention, 2014, 9(2)), realized the quantitative analysis of the microstructure of TC4 titanium alloy through image enhancement, segmentation, thresholding and other methods. Sun Chaoming et al., in "Progress in Metallographic Grain Size Quantitative Assessment Technology" (Physical and Chemical Testing - Physics, 2012, 48(12): 814-817), analyzed the key technologies in metallographic grain size quantitative assessment, proposed the problem of accurate extraction and reconstruction of grain boundaries in complex metallographic images, and analyzed the progress of metallographic grain size assessment technology at home and abroad. From the research situation in the industry, it can be seen that the current research on metallographic automatic quantitative detection is mainly based on traditional image processing, thresholding and other technologies or single variety types. The problems of large fluctuations in detection results and the need for manual intervention have not been well solved. Summary of the invention

[0003] Based on the above purpose, the present invention proposes a method for intelligent grading of steel grain size, comprising: Obtaining and preparing the steel to be rated to obtain a grain size sample; Scanning and collecting the field of view of the grain size sample to obtain a metallographic picture; Processing the metallographic image through a preset intelligent rating model to obtain a grain size rating; The grain size rating result is subjected to an accuracy test to obtain the accuracy, and the grain size rating and the accuracy are output.

[0004] In some embodiments, the step of obtaining and preparing the steel to be rated to obtain a grain size sample includes: Processing the steel to be rated into a preset size by an automatic sample mounting machine to obtain a sample; According to the tensile strength index of the sample, an automatic grinding and polishing machine is used to cut the sample, and abrasives and polishing cloths corresponding to the tensile strength index of the sample are selected to perform coarse grinding, fine grinding, coarse polishing and fine polishing in sequence to obtain a grain size sample.

[0005] In some embodiments, the preset intelligent rating model construction process includes: Obtain grain size samples, perform preprocessing, and divide them into a grain size training set and a grain size test set; Extracting grain size features and corresponding rating labels from the grain size training set; Inputting the grain size characteristics and the corresponding rating labels into a preset multi-stage global trainable neural network model for training, and obtaining a grain size rating model when the number of iterations reaches a preset number; The grain size rating model is verified through the grain size test set, and the verified grain size rating model is used as the intelligent rating model.

[0006] In some embodiments, the step of obtaining a grain size sample and performing preprocessing includes: Obtain grain size samples, collect original grain size metallographic images after marking, grayscale, binarize and denoise the image data to obtain processed data; The processed data is marked, the grains are annotated along the grain boundaries, and the images before and after marking are imported into the corresponding database for grain size feature extraction to obtain the grain size metallographic rating label.

[0007] In some embodiments, the process of inputting the grain size features and the corresponding rating labels into a preset multi-stage global trainable neural network model for training, and obtaining the grain size rating model when the number of iterations reaches a preset number, includes: The grain size characteristics are input into a preset multi-stage global trainable neural network model for multi-stage training.

[0008] Calculating a loss function based on the output and the rating label; The gradient of the loss function to the model parameters is calculated through the back-propagation algorithm, and the model parameters are iteratively updated using the gradient descent algorithm; When the number of iterations reaches the preset number, the grain size rating model is obtained.

[0009] In some embodiments, the grain size rating result is subjected to an accuracy test, and the process of obtaining the accuracy includes: Select the pass rate as the accuracy indicator according to the test standard; The pass rate of grain size rating is calculated by random sampling.

[0010] In some embodiments, the method further comprises: After the first intelligent grain size rating, the intelligent grain size rating is performed again at a fixed interval; Compare the two rating results to get the deviation; The repeatability was calculated based on the distribution of the deviations.

[0011] The present invention proposes a system for intelligently grading steel grain size, comprising: A preparation unit is configured to obtain and prepare the steel to be graded to obtain a grain size sample; A collection unit configured to scan and collect the field of view of the grain size sample to obtain a metallographic image; A rating unit configured to process the metallographic image through a preset intelligent rating model to obtain a grain size rating; The testing unit is configured to perform an accuracy test on the grain size rating result, obtain the accuracy, and output the grain size rating and the accuracy.

[0012] The present invention provides a computer device, comprising: At least one processor; and a memory, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the program, the steps of the method for intelligent grading of steel grain size are performed.

[0013] The present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for intelligent grading of steel grain size are performed.

[0014] The present invention has at least the following beneficial technical effects: The present invention proposes a method, system, equipment and medium for intelligent rating of steel grain size. The method comprises: obtaining steel to be rated and preparing it to obtain a grain size sample; scanning and collecting the field of view of the grain size sample to obtain a metallographic image; processing the metallographic image through a preset intelligent rating model to obtain a grain size rating; performing an accuracy test on the grain size rating result to obtain the accuracy, and outputting the grain size rating and accuracy.

[0015] The present invention improves the quality of grain size sample preparation, based on artificial intelligence and image recognition technology, and uses convolutional neural network deep learning and other methods to establish an intelligent grain size rating model. The metallographic sample is automatically scanned and the field of view is automatically collected through an automatic metallographic microscope. No less than 300 metallographic pictures are collected for each sample. The rating model is used to rate each picture by comparison method, and 5 representative fields of view closest to the mean are selected. The corresponding pictures are retained and graded according to the average value. It has the characteristics of fast detection speed, low manual dependence, and good repeatability of detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other embodiments can be obtained based on these drawings without paying creative work.

[0017] Figure 1 A flow chart of a method for intelligent grading of steel grain size provided by the present invention; Figure 2 A system module diagram of intelligent steel grain size rating provided by the present invention; Figure 3 A detection flow chart of a method for intelligent grain size rating of steel materials according to an embodiment of the present invention; Figure 4 An original picture collected by a microscope in one embodiment of a method for intelligent grading of steel grain size provided by the present invention; Figure 5 This is a pre-processed picture of an embodiment of a method for intelligent grading of steel grain size provided by the present invention; Figure 6 This is a LabelMe manual marking diagram of an embodiment of a method for intelligent grading of steel grain size provided by the present invention; Figure 7 A metallographic label diagram of a grain size of an embodiment of a method for intelligent grain size rating of steel provided by the present invention; Figure 8 A schematic diagram of a neural network learning model structure of an embodiment of a method for intelligent rating of steel grain size provided by the present invention; Fig. 9 A trend chart of the qualified rate and sample size of the test results of an embodiment of a method for intelligent grading of steel grain size provided by the present invention; Fig.10 A trend diagram of the repeatability of test results and the sample size of an embodiment of a method for intelligent grading of steel grain size provided by the present invention; Fig.11 A diagram showing the effect of using the method for intelligently grading grain size of steel provided by the present invention; Fig.12 A schematic diagram of the structure of an embodiment of a computer device provided by the present invention; Fig.13 A schematic diagram of the structure of an embodiment of a computer-readable storage medium provided by the present invention. DETAILED DESCRIPTION

[0018] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the following further elaborates on the embodiments of the present invention in detail with reference to specific embodiments and the accompanying drawings.

[0019] It should be noted that all the expressions using "first" and "second" in the embodiments of the present invention are for distinguishing two entities or parameters with the same name but different identities. It can be seen that "first" and "second" are only for the convenience of expression and should not be construed as a limitation on the embodiments of the present invention. This will not be elaborated one by one in the subsequent embodiments.

[0020] The present invention proposes a method for intelligent grading of the grain size of steel. Please refer to Figure 1 and Figure 3 , including: S1: Obtain the steel to be graded and perform preparation to obtain a grain size specimen; S2: Scan and collect fields of view of the grain size specimen to obtain metallographic pictures; S3: Process the metallographic pictures through a preset intelligent grading model to obtain a grain size grade; S4: Perform an accuracy test on the grain size grading result to obtain the accuracy, and output the grain size grade and the accuracy.

[0021] The present invention uses the neural network deep learning method to automatically evaluate the grain size of steel. By manually marking the grain size samples and other methods, a grain size feature database is established, and the computer artificial intelligence and neural network deep learning technologies are used to construct a standard model of the grain size convolutional neural network. Then, a specified sample preparation method is used to ensure the sample preparation quality, and an automatic metallographic microscope is used to automatically scan and collect fields of view of the metallographic specimens. Each specimen collects no less than 300 metallographic pictures. The grading model uses the comparison method to grade each picture one by one, selects 5 representative fields of view closest to the average value, saves the corresponding pictures and grades them according to the average value, and finally gives the grain size result. A set of intelligent grading methods for the grain size of steel based on artificial intelligence is developed, with fast detection speed, low dependence on manual labor, and good repeatability of detection results.

[0022] Complex metallographic tissue images cannot perform the automatic evaluation of grain size through simple image processing, segmentation, thresholding, etc. Therefore, based on image preprocessing, edge segmentation, thresholding, and morphology, the present invention invented an intelligent grading method for grain size based on adaptive manual marking. The main process of the method is divided into three steps (see in detail Figure 3): ① Sample preparation: automatic grinding and polishing are used to prepare the samples, and the parallelism of the samples is required to be ≤0.05mm, and the test surface is free of scratches and polishing strain marks; ② Intelligent rating model grading: the grain size samples are scanned and the field of view is collected through an automatic metallographic microscope. No less than 300 metallographic images are collected for each sample. The rating model uses the comparison method to grade each sample one by one, and automatically retains 5 representative field of view images closest to the mean for traceability, and the grade is determined according to the average value; ③ The results are reported.

[0023] In some embodiments, see Figure 1 The step of obtaining the steel to be rated and preparing it to obtain a grain size sample includes: Processing the steel to be rated into a preset size by an automatic sample mounting machine to obtain a sample; According to the tensile strength index of the sample, an automatic grinding and polishing machine is used to cut the sample, and abrasives and polishing cloths corresponding to the tensile strength index of the sample are selected to perform coarse grinding, fine grinding, coarse polishing and fine polishing in sequence to obtain a grain size sample.

[0024] Intelligent grain size rating has high requirements on the preparation quality of the sample. The parallelism between the detection surface and the opposite surface is required to be ≤0.05mm (according to the automatic focusing stroke setting of the automatic metallographic microscope), and the detection surface must not have defects such as scratches and polishing strain marks, otherwise it will affect the automatic focusing and scanning framing of the microscope. The sample is prepared by the traditional metallographic pre-grinder and polisher method. The parallelism is generally 0.10-0.15mm. Affected by manual experience, the detection surface has scratches and polishing strain marks of varying degrees. The quality of sample preparation is difficult to meet the requirements of intelligent rating for sample preparation quality. The present invention uses an automatic mounting machine and an automatic grinding and polishing machine for sample preparation. According to the tensile strength index of the sample, a set of automatic grinding and polishing operation methods adapted to different product characteristics are formulated (see Table 1).

[0025] Table 1 Automatic grinding and polishing method for intelligent grain size assessment samples

[0026] Using an automatic mounting machine can improve the efficiency and accuracy of sample preparation and reduce errors caused by human factors. At the same time, it can also ensure the stability and consistency of the sample during the mounting process, laying a good foundation for subsequent grinding and polishing operations. The automatic grinding and polishing machine can run uninterruptedly, significantly improving the grinding and polishing efficiency, and is especially suitable for processing large quantities of samples. This shortens the time for sample preparation and improves the overall work efficiency of the laboratory.

[0027] In some embodiments, see Figure 1 and Figure 8 , the preset intelligent rating model construction process includes: Obtain grain size samples, perform preprocessing, and divide them into a grain size training set and a grain size test set; Extracting grain size features and corresponding rating labels from the grain size training set; Inputting the grain size characteristics and the corresponding rating labels into a preset multi-stage global trainable neural network model for training, and obtaining a grain size rating model when the number of iterations reaches a preset number; The grain size rating model is verified through the grain size test set, and the verified grain size rating model is used as the intelligent rating model.

[0028] The rating model mainly includes: manual marking of grain size samples and establishment of an image database; inputting the database into a convolutional neural network model for deep learning, extracting features, and establishing an automatic grain size rating model; testing, training, adapting and optimizing the model to reduce the difference between the method's detection results and manual ones and improve repeatability.

[0029] The grain size rating model is mainly a convolutional neural network model built using computer artificial intelligence and neural network deep learning technology. Its core is a multi-stage globally trainable artificial neural network learning model that can build different network structures for different grain sizes and morphologies, and directly learn high-level and abstract features from the data set. The typical neural network model structure consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. For a detailed structural diagram, see Figure 8 In the above figure, the input layer is mainly used for the input of original samples. In order to reduce the complexity of subsequent algorithm processing, this paper uses grayscale to process images; the convolution layer is used for image feature extraction, mapping and in-depth understanding of learning, mapping from one plane to the next plane, and automatically extracting local features of the image through neurons. Each neuron is connected to the local receptive field of the previous layer. The neurons of each plane in each layer extract local features of specific areas in the image, such as directional features, morphological features, edge features, etc.; the pooling layer is a layer for calculating local averages and secondary feature extraction. Through the secondary feature extraction structure, the convolutional neural network has a higher distortion tolerance for input samples during recognition and detects more feature information; the fully connected layer is used to connect and integrate all features, and send the output value to the classifier to map it to the sample label.

[0030] Through the convolutional neural network model, the low-latitude, mid-latitude, and high-latitude features of each level of grain size are extracted from the sample database, and then classified according to the grain size level to establish an intelligent grain size rating model.

[0031] After the model is initially established, there are certain fluctuations and errors in its rating results. It is necessary to use unlabeled grain size samples for training and optimization to reduce detection errors and improve detection repeatability.

[0032] By training with a multi-stage globally trainable neural network model, the model is able to learn the complex relationship between grain size features and rating labels, thereby achieving accurate rating of grain size. The neural network model has a strong nonlinear mapping capability, can process complex input data, and output accurate rating results. Once the intelligent rating model is trained, it can quickly rate new grain size samples, greatly improving the rating efficiency. Compared with traditional manual rating methods, the intelligent rating model can significantly reduce the time and labor costs required for rating. Since the intelligent rating model is learned based on training data, it can ensure consistent rating of grain size samples from different batches and sources. This helps to maintain the stability and reliability of the rating results.

[0033] In some embodiments, see Figure 1 , Figure 4 , Figure 5 , Figure 6 and Figure 7 The steps for obtaining grain size samples and preprocessing include: Obtain grain size samples, collect original grain size metallographic images after marking, grayscale, binarize and denoise the image data to obtain processed data; The processed data is marked, the grains are annotated along the grain boundaries, and the images before and after marking are imported into the corresponding database for grain size feature extraction to obtain the grain size metallographic rating label.

[0034] Manual labeling and image database establishment The database is a series of important data sets used for automatic analysis and processing of specific metallographic images. It is necessary to extract the original features through manual marking and annotation of grain size metallographic images, establish data sets at various levels, and form a database of specific features when the sample size reaches a certain level, and finally realize automatic analysis of grain size.

[0035] The original grain size metallographic image was obtained using a metallographic microscope (see Figure 4 ), after grayscale processing, the pre-processed metallographic image sample is obtained (see Figure 5 ).

[0036] The deep learning labeling tool LabelMe is used to manually label the pre-processed metallographic images. When labeling, the grains need to be labeled along the grain boundaries. Some fuzzy grain boundaries can be ignored during the initial labeling, but the labeled ones must be clear and accurate to facilitate the subsequent accurate extraction of grain features. The manual labeling results are shown in Figure 6 .

[0037] After completing manual marking, the images before and after marking are imported into the database of the corresponding level for grain size feature extraction to generate the grain size metallographic label of the sample (see Figure 7), Figure 7 When generating metallographic labels, only the marking marks are retained and the entire background needs to be shielded, so no scale is set in the image.

[0038] By importing the images before and after marking into the database and using advanced image processing technology, the characteristics of grain size can be accurately extracted. These characteristics include the size, shape, and distribution of the grains, which are the key basis for rating grain size. Accurate feature extraction helps reduce rating errors caused by human factors and improve the accuracy of rating. After establishing a database containing grain size characteristics at all levels, a rating model can be trained using machine learning or deep learning technology. The model can automatically identify and rate new grain size images, thereby achieving automated rating. Automated rating can significantly improve rating efficiency and reduce the time and labor costs required for rating.

[0039] In some embodiments, see Figure 1 The process of inputting the grain size characteristics and the corresponding rating labels into a preset multi-stage global trainable neural network model for training, and obtaining the grain size rating model when the number of iterations reaches a preset number, includes: The grain size characteristics are input into a preset multi-stage global trainable neural network model for multi-stage training.

[0040] Calculating a loss function based on the output and the rating label; The gradient of the loss function to the model parameters is calculated through the back-propagation algorithm, and the model parameters are iteratively updated using the gradient descent algorithm; When the number of iterations reaches the preset number, the grain size rating model is obtained.

[0041] Through continuous iterative training, the neural network model can learn the complex mapping relationship between grain size characteristics and rating labels. This mapping relationship not only takes into account intuitive features such as grain size and shape, but may also involve deeper features such as texture and distribution. During the training process, the model will continuously optimize its internal parameters based on the feedback of the training data, thereby improving the accuracy of the rating.

[0042] The multi-stage globally trainable neural network model can continuously adjust and optimize the parameters of each stage during the entire training process to ensure that the entire model reaches the optimal state as a whole. This global optimization capability helps to improve the stability and reliability of the rating model and reduce rating errors caused by improper model parameters.

[0043] Since the neural network model has strong nonlinear fitting capabilities, it is able to handle various complex grain size characteristics and rate different types of materials. This enhances the generalization ability of the model and enables it to be applied to a wider range of scenarios.

[0044] In some embodiments, see Fig. 9 The grain size rating result is subjected to an accuracy test, and the process of obtaining the accuracy includes: Select the pass rate as the accuracy indicator according to the test standard; The pass rate of grain size rating is calculated by random sampling.

[0045] The present invention uses a multi-station automatic metallographic microscope for training and optimization, and the content is mainly divided into two aspects. One is the comparison between the model and manual detection: the grain size sample is used as the material for testing, and the samples to be tested are manually placed on the multi-station automatic stage in turn, and the sample code, detection items and other information are edited in sequence according to the station number. The microscope automatically identifies the sample, scans, frames, and rates the rating model, and outputs the grain size grade number (the result retains two decimal places). The output result is compared with the manual detection result. When the detection result deviation is less than or equal to 0.5 levels, the comparison result is qualified.

[0046] More than 300 images were processed, and samples with deviations between the model and manual inspection results exceeding the expected range were manually marked separately, imported into the convolutional neural network model, and the standard library was optimized. The correlation between the qualified rate, repeatability and sample quantity of each grain size result comparison is shown in Table 2 Table 2 Statistics of sample size and result deviation

[0047] from Fig. 9 It can be seen that with the increase in the number of samples in the standard library, the pass rate of the rating model detection results and manual comparison has gradually improved. The continuous and stable addition of new sample features to the convolutional neural network model is the guarantee for the grain size intelligent rating model to output more accurate quantitative analysis results.

[0048] Accuracy testing verifies the accuracy of rating results by comparing them with reference standards of known accuracy or actual observations. This helps ensure the authenticity and reliability of rating results and avoids decision-making errors caused by rating errors. Through continuous testing and optimization, the performance of rating models and algorithms can be gradually improved, enabling them to more accurately identify and rate grain size features, thereby improving the accuracy and efficiency of ratings.

[0049] In some embodiments, see Fig.10 , the method further comprises: After the first intelligent grain size rating, the intelligent grain size rating is performed again at a fixed interval; Compare the two rating results to get the deviation; The repeatability was calculated based on the distribution of the deviations.

[0050] The second is the repeatability of the model rating results: the grain size sample is used as the material for testing. After the first automatic rating is completed, the sample order on the automatic stage is mixed up or mixed with other samples and rearranged, and automatic rating is carried out again. When the deviation of the two results of the same sample is less than or equal to 0.25 levels, the model rating results are considered to have good repeatability.

[0051] from Fig.10 It can be seen that with the increase in the number of samples in the standard library, the repeatability of the rating model detection results is gradually improved. The continuous and stable addition of new sample features to the convolutional neural network model is the guarantee for the grain size intelligent rating model to output more accurate quantitative analysis results.

[0052] If there is a significant deviation between the two rating results, it may mean that there are some potential problems in the rating system, such as sensor failure, data processing errors, or algorithm defects. This helps to detect and fix these problems in a timely manner to ensure the accuracy and reliability of the rating system.

[0053] By repeating the rating and calculating the repeatability, the reliability and consistency of the rating results can be verified. If the two rating results are highly consistent, it means that the rating system has high stability and accuracy, thus enhancing the credibility of the rating results. Repeatability testing can be used as an important indicator to evaluate the performance of the rating algorithm. If the repeatability is low, it means that the algorithm may have large fluctuations when processing the same or similar samples, which helps to identify deficiencies in the algorithm and optimize it.

[0054] The present invention proposes a system for intelligent grading of steel grain size. Figure 2 ,include: The preparation unit 100 is configured to obtain and prepare the steel to be graded to obtain a grain size sample; A collection unit 200 is configured to scan and collect the field of view of the grain size sample to obtain a metallographic image; The rating unit 300 is configured to process the metallographic image through a preset intelligent rating model to obtain a grain size rating; The testing unit 400 is configured to perform an accuracy test on the grain size rating result, obtain the accuracy, and output the grain size rating and the accuracy.

[0055] In some embodiments, see Fig.11 , randomly selected some samples for grinding and polishing according to the grain size intelligent assessment sample preparation method formulated in the article, the parallelism of the samples was measured using an automatic measuring table with a measurement accuracy of 0.001mm, the surface quality of the field of view was observed through a metallographic microscope, and the preparation effect of the samples was evaluated. A total of 70 samples were randomly checked.

[0056] Table 3 Statistical table of sample preparation effects

[0057] It can be seen from Table 3 above that the grain size specimens prepared by the grain size intelligent assessment specimen preparation method have good parallelism, clear grain boundary corrosion, and the specimen preparation quality meets the intelligent rating requirements.

[0058] A certain number of samples were randomly selected for intelligent grain size rating and manual inspection. The number of samples in each category was required to be no less than 30. A total of 43 steel grades and 600 samples were randomly selected. According to the provisions on reproducibility in GB / T6394-2017 "Method for Determination of Average Grain Size of Metals", the deviation between the intelligent rating result and the manual inspection result was set to be qualified if it was within 0.5 levels. The qualified rate of the intelligent grain size rating results reached 98.2%, as shown in Table 4 for details.

[0059] Table 4 Statistics of deviations between intelligent rating and manual detection results

[0060] While conducting intelligent rating and manual comparison, a repeatability test of the grain size intelligent rating results was carried out simultaneously: within 8 hours after the first intelligent grain size rating, the intelligent grain size rating was carried out again, and the two results were compared. According to the repeatability requirements of GB / T6394, the repeatability is good if the deviation is within 0.25 levels. The repeatability of the grain size intelligent rating results reached 98.8%, see Table 5 for details.

[0061] Table 5 Statistics of repeatability of intelligent rating results

[0062] According to the grain size intelligent rating method detection process, 32 sets of grain size samples were randomly prepared and divided into 4 groups. Four personnel with ≥5 years of work experience and rich metallographic inspection experience were organized to carry out grain size rating work according to manual and intelligent processes respectively. The time taken for each process was statistically analyzed. See Table 6, Table 7, Fig.11 .

[0063] Table 6 Statistics of sample preparation efficiency improvement of grain size intelligent rating method

[0064] Table 7 Statistics on improvement of detection efficiency of grain size intelligent rating method

[0065] From above Fig.11 It can be seen that compared with the traditional manual method, the sample preparation efficiency of the intelligent grain size rating method of the present invention is increased by 23.8% and the grain size detection efficiency is increased by 41.1%.

[0066] Based on the same inventive concept, according to another aspect of the present invention, Fig.12 As shown, an embodiment of the present invention further provides a computer device 30, which includes a processor 310 and a memory 320. The memory 320 stores a computer program 321 that can be run on the processor. When the processor 310 executes the program, the steps of the above method are performed.

[0067] Based on the same inventive concept, according to another aspect of the present invention, Fig.13 As shown, an embodiment of the present invention further provides a computer-readable storage medium 40, which stores a computer program 410 for executing the above method when executed by a processor.

[0068] The embodiment of the present invention may also include a corresponding computer device. The computer device includes a memory, at least one processor, and a computer program stored in the memory and executable on the processor, and the processor executes any one of the above methods when executing the program.

[0069] The memory, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions / modules in the embodiments of the present application. The processor executes various functional applications and data processing of the device by running the non-volatile software programs, instructions and modules stored in the memory, that is, implementing the above method.

[0070] The memory may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the device, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In an embodiment, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the local module via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0071] Finally, it should be noted that a person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium of the program can be a disk, an optical disk, a read-only storage memory (ROM) or a random access memory (RAM), etc. The above-mentioned computer program embodiments can achieve the same or similar effects as the corresponding above-mentioned arbitrary method embodiments.

[0072] It will also be appreciated by those skilled in the art that various exemplary logic blocks, modules, circuits and algorithm steps described in conjunction with the disclosure herein can be implemented as electronic hardware, computer software or a combination of the two. In order to clearly illustrate this interchangeability of hardware and software, a general description has been given to the functions of various schematic components, blocks, modules, circuits and steps. Whether this function is implemented as software or hardware depends on specific applications and the design constraints imposed on the entire system. Those skilled in the art can implement the function in various ways for each specific application, but this implementation decision should not be interpreted as causing a departure from the disclosed scope of the embodiments of the present invention.

[0073] The above are exemplary embodiments disclosed in the present invention, but it should be noted that various changes and modifications may be made without departing from the scope of the embodiments disclosed in the present invention as defined in the claims. The functions, steps and / or actions of the method claims according to the disclosed embodiments described herein do not need to be performed in any particular order. The serial numbers of the embodiments disclosed in the above embodiments of the present invention are for description only and do not represent the advantages and disadvantages of the embodiments. In addition, although the elements disclosed in the embodiments of the present invention may be described or required in individual form, they may also be understood as multiple unless explicitly limited to the singular.

[0074] It should be understood that, as used herein, the singular forms "a", "an" are intended to include the plural forms as well, unless the context clearly supports an exception. It should also be understood that, as used herein, "and / or" refers to any and all possible combinations including one or more of the associated listed items.

[0075] A person skilled in the art should understand that the discussion of any of the above embodiments is only exemplary and is not intended to imply that the scope of the disclosure of the embodiments of the present invention (including the claims) is limited to these examples; under the concept of the embodiments of the present invention, the technical features in the above embodiments or different embodiments can also be combined, and there are many other changes in different aspects of the above embodiments of the present invention, which are not provided in detail for the sake of simplicity. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present invention should be included in the protection scope of the embodiments of the present invention.

Claims

1. A method for intelligent grading of steel grain size, characterized in that: include: Obtaining and preparing the steel to be rated to obtain a grain size sample; Scanning and collecting the field of view of the grain size sample to obtain a metallographic image; Processing the metallographic image through a preset intelligent rating model to obtain a grain size rating; The grain size rating result is subjected to an accuracy test to obtain the accuracy, and the grain size rating and the accuracy are output.

2. The method for intelligent grading of steel grain size according to claim 1, characterized in that: The step of obtaining the steel to be rated and preparing it to obtain a grain size sample comprises: Processing the steel to be rated into a preset size by an automatic sample mounting machine to obtain a sample; According to the tensile strength index of the sample, an automatic grinding and polishing machine is used to cut the sample, and abrasives and polishing cloths corresponding to the tensile strength index of the sample are selected to perform coarse grinding, fine grinding, coarse polishing and fine polishing in sequence to obtain a grain size sample.

3. The method for intelligent grading of steel grain size according to claim 1, characterized in that: The preset intelligent rating model construction process includes: Obtain grain size samples, perform preprocessing, and divide them into a grain size training set and a grain size test set; Extracting grain size features and corresponding rating labels from the grain size training set; Inputting the grain size characteristics and the corresponding rating labels into a preset multi-stage global trainable neural network model for training, and obtaining a grain size rating model when the number of iterations reaches a preset number; The grain size rating model is verified through the grain size test set, and the verified grain size rating model is used as the intelligent rating model.

4. The method for intelligent grading of steel grain size according to claim 1, characterized in that: The steps for obtaining grain size samples and preprocessing include: Obtain grain size samples, collect original grain size metallographic images after marking, grayscale, binarize and denoise the image data to obtain processed data; The processed data is marked, the grains are annotated along the grain boundaries, and the images before and after marking are imported into the corresponding database for grain size feature extraction to obtain the grain size metallographic rating label.

5. The method for intelligent grading of steel grain size according to claim 3, characterized in that: The process of inputting the grain size characteristics and the corresponding rating labels into a preset multi-stage global trainable neural network model for training, and obtaining the grain size rating model when the number of iterations reaches a preset number, comprises: Inputting grain size characteristics into a preset multi-stage global trainable neural network model for multi-stage training; Calculating a loss function based on the output and the rating label; The gradient of the loss function to the model parameters is calculated through the back-propagation algorithm, and the model parameters are iteratively updated using the gradient descent algorithm; When the number of iterations reaches the preset number, the grain size rating model is obtained.

6. The method for intelligent grading of steel grain size according to claim 1, characterized in that: The grain size rating result is subjected to accuracy testing, and the process of obtaining the accuracy includes: Select the pass rate as the accuracy indicator according to the test standard; The pass rate of grain size rating is calculated by random sampling.

7. The method for intelligently grading steel grain size according to claim 1, characterized in that: The method also includes: After the first intelligent grain size rating, the intelligent grain size rating is performed again at a fixed interval; Compare the two rating results to get the deviation; The repeatability was calculated based on the distribution of the deviations.

8. A system for intelligent grading of steel grain size, characterized in that: include: A preparation unit is configured to obtain and prepare the steel to be graded to obtain a grain size sample; A collection unit configured to scan and collect the field of view of the grain size sample to obtain a metallographic image; A rating unit configured to process the metallographic image through a preset intelligent rating model to obtain a grain size rating; The testing unit is configured to perform an accuracy test on the grain size rating result, obtain the accuracy, and output the grain size rating and the accuracy.

9. A computer device comprising: at least one processor; and a memory storing a computer program executable on the processor, wherein the processor executes the steps of a method for intelligent grading of steel grain size as claimed in any one of claims 1 to 7 when executing the program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for intelligent grading of steel grain size as claimed in any one of claims 1 to 7 are performed.

Images